Head-to-head comparison

GPT-6 Astra vs. DeepSeek-V4.1-Flash

Compare published benchmark results from matching versions and cohorts, API pricing, and technical specifications. Any documented test setup differences remain visible.

shared result series: 214
distinct benchmarks: 26
latest retrieval: 2026-10-15

First modelSecond model

Specifications

General specifications

Release date, availability, context window, architecture, and published parameter counts at a glance. Undisclosed details remain marked as unavailable.

Specifications for GPT-6 Astra and DeepSeek-V4.1-Flash
Attribute	GPT-6 Astra	DeepSeek-V4.1-Flash
Provider	OpenAI	DeepSeek
Released	Sep 3, 2026	Sep 10, 2026
Availability	Limited access	Active
Model type	Proprietary	Open Weights
Parameters	Not published	552B
Architecture	Not published	Mixture of Experts
Context window	1,050,000 tokens	1,000,000 tokens
Knowledge cutoff	April 30, 2026	Not published
Technical sources	Model, Context	Model

Pricing

API pricing compared

The table compares official standard API rates per 1M tokens, including published cache pricing. Batch, fast, priority, regional surcharges, tool calls, and cloud platform pricing are excluded.

API pricing for GPT-6 Astra and DeepSeek-V4.1-Flash
Price type	GPT-6 Astra	DeepSeek-V4.1-Flash
API input	$10 ≤272K / $20 >272K	$0.15 (Off-Peak)
API output	$50 ≤272K / $75 >272K	$0.6 (Off-Peak)
Cache read	$1 ≤272K / $2 >272K	$0.003 (Off-Peak)
Cache write (5 min.)	$12.5 ≤272K / $25 >272K	Not listed
Cache write (1 hr.)	Not listed	Not listed
Price verified	09/22/2026 Source	09/23/2026 Source
DeepSeek-V4.1-Flash: peak pricing is $0.3 input / $1.2 output per 1M tokens during Mon, Tue, Wed, Thu, Fri, 01:00-04:00 UTC; Mon, Tue, Wed, Thu, Fri, 06:00-10:00 UTC; off-peak rates apply at all other times.

GPT-6 AstraDeepSeek-V4.1-Flash

API inputUSD per 1M tokens, base tier

GPT-6 Astra$10

DeepSeek-V4.1-Flash$0.15

API outputUSD per 1M tokens, base tier

GPT-6 Astra$50

DeepSeek-V4.1-Flash$0.6

Cache readUSD per 1M tokens, base tier

GPT-6 Astra$1

DeepSeek-V4.1-Flash$0.003

Key to the marks:better valuebetter, but not conclusivetie

Pricing by provider

Direct developer pricing and provider endpoint tariffs routed through OpenRouter are listed separately. Prices are in USD per 1M tokens. Context-dependent direct tiers plus regional, Flex, Priority, and other OpenRouter tariffs remain individually identifiable.

Provider pricing for GPT-6 Astra and DeepSeek-V4.1-Flash
Model	Provider	Purchase route and tariff	Input per 1M tokens	Output per 1M tokens	Source
GPT-6 Astra	OpenAI	Direct from developerUp to 272,000 context tokens	$10	$50	Source09/22/2026
	OpenAI	Direct from developerAbove 272,000 context tokens	$20	$75	Source09/22/2026
	Amazon Bedrock	Via OpenRouteramazon-bedrock/us-west-2	$11	$55	Source10/17/2026
	Azure	Via OpenRouterazure	$10	$50	Source10/17/2026
	Azure	Via OpenRouterazure/us	$11	$55	Source10/17/2026
	OpenAI	Via OpenRouteropenai	$10	$50	Source10/17/2026
	OpenAI	Via OpenRouteropenai/fast	$20	$100	Source10/17/2026
	OpenAI	Via OpenRouteropenai/flex	Lowest listed price: $5	Lowest listed price: $25	Source10/17/2026
	OpenAI	Via OpenRouteropenai/ultrafast	$60	$300	Source10/17/2026
DeepSeek-V4.1-Flash	DeepSeek	Direct from developerStandard API	$0.15	$0.6	Source09/23/2026
	Alibaba	Via OpenRouteralibaba	$0.3	$1.2	Source10/17/2026
	AtlasCloud	Via OpenRouteratlas-cloud/fp8	$0.135	$0.54	Source10/17/2026
	Baidu	Via OpenRouterbaidu/fp8	$0.3	$1.2	Source10/17/2026
	BaseTen	Via OpenRouterbaseten	$0.3	$1.2	Source10/17/2026
	BaseTen	Via OpenRouterbaseten/us	$0.3	$1.2	Source10/17/2026
	Cloudflare	Via OpenRoutercloudflare	$0.1	$0.9	Source10/17/2026
	CoreWeave	Via OpenRoutercoreweave/fp8	$0.15	$0.65	Source10/17/2026
	Crusoe	Via OpenRoutercrusoe/fp8	$0.29	$1.2	Source10/17/2026
	Decart	Via OpenRouterdecart/fp4	$0.09	$0.18	Source10/17/2026
	DeepInfra	Via OpenRouterdeepinfra/fp8	$0.14	$0.42	Source10/17/2026
	DeepInfra	Via OpenRouterdeepinfra/fp8/flex	$0.112	$0.336	Source10/17/2026
	DeepInfra	Via OpenRouterdeepinfra/fp8/priority	$0.21	$0.63	Source10/17/2026
	DeepSeek	Via OpenRouterdeepseek	$0.15	$0.6	Source10/17/2026
	DigitalOcean	Via OpenRouterdigitalocean	$0.195	$0.78	Source10/17/2026
	Fireworks	Via OpenRouterfireworks	$0.3	$1.2	Source10/17/2026
	Fireworks	Via OpenRouterfireworks/us	$0.45	$1.8	Source10/17/2026
	GMICloud	Via OpenRoutergmicloud/fp8	$0.18	$0.72	Source10/17/2026
	InferenceNet	Via OpenRouterinference-net/fp8	$0.103	$0.715	Source10/17/2026
	Ionstream	Via OpenRouterionstream	$0.27	$1.1	Source10/17/2026
	Makora	Via OpenRoutermakora/fp8	$0.27	$1.15	Source10/17/2026
	Modal	Via OpenRoutermodal	$0.3	$1.2	Source10/17/2026
	Morph	Via OpenRoutermorph/fp8	$0.04	$0.7	Source10/17/2026
	Novita	Via OpenRouternovita/fp8	$0.195	$0.78	Source10/17/2026
	OpenInference	Via OpenRouteropen-inference/fp4	Lowest listed price: $0.032	Lowest listed price: $0.144	Source10/17/2026
	Parasail	Via OpenRouterparasail/fp8	$0.3	$1.2	Source10/17/2026
	Phala	Via OpenRouterphala	$0.21	$0.84	Source10/17/2026
	PrimeIntellect	Via OpenRouterprimeintellect	$0.3	$1.2	Source10/17/2026
	Relace	Via OpenRouterrelace	$0.2	$0.6	Source10/17/2026
	Sail Research	Via OpenRoutersail-research/fp4	$0.09	$0.3	Source10/17/2026
	SiliconFlow	Via OpenRoutersiliconflow/fp8	$0.3	$1.2	Source10/17/2026
	StreamLake	Via OpenRouterstreamlake/fp8	$0.1497	$0.5988	Source10/17/2026
	Together	Via OpenRoutertogether	$0.3	$1.2	Source10/17/2026
	Venice	Via OpenRoutervenice/fp8	$0.3	$1.2	Source10/17/2026
	Wafer	Via OpenRouterwafer/mxfp4	$0.1	$0.65	Source10/17/2026

Performance

Shared benchmarks

Only published results with the same benchmark version, task, metric, and comparison cohort are paired. Different reasoning levels, agents, harnesses, or output limits appear directly in the table.

Capability profile from matched benchmarks

Each axis averages directly matched benchmark families. A value of 100 means the stronger value within this pair, not a universal quality score.

GPT-6 Astra

DeepSeek-V4.1-Flash

Coding: 100 / 95 (2 benchmark families)

Reasoning: 99.8 / 99.9 (1 benchmark family)

Mathematics: 99.5 / 100 (1 benchmark family)

Knowledge: 100 / 99.5 (1 benchmark family)

Finance: 99.7 / 100 (1 benchmark family)

Legal: 100 / 97.9 (1 benchmark family)

Multimodal: 100 / 97.7 (1 benchmark family)

Health: 97.4 / 100 (1 benchmark family)

40 matched rows from 3 benchmark families were considered. The radar shows 8 of 9 comparable categories.

GPT-6 AstraDeepSeek-V4.1-Flash

BioMysteryBenchOverall. Higher is better. Test setup not documented as identical

GPT-6 Astra79.259 %

DeepSeek-V4.1-Flash67.778 %

Code MigrationOverall. Higher is better. Test setup not documented as identical

GPT-6 Astra67.74 %

DeepSeek-V4.1-Flash45.623 %

CyberBench v1.1Overall. Higher is better. Test setup not documented as identical

GPT-6 Astra41.072 %

DeepSeek-V4.1-Flash73.69 %

EMBOverall. Higher is better. Test setup not documented as identical

GPT-6 Astra71.7 %

DeepSeek-V4.1-Flash57.208 %

Finance Agent (v2)Overall. Higher is better. Test setup not documented as identical

GPT-6 Astra53.54 %

DeepSeek-V4.1-Flash53.481 %

Harvey's Legal Agent BenchmarkOverall · Task fully resolved. Higher is better. Test setup not documented as identical

GPT-6 Astra5.417 %

DeepSeek-V4.1-Flash6.667 %

IOIOverall. Higher is better. Test setup not documented as identical

GPT-6 Astra100 %

DeepSeek-V4.1-Flash40.278 %

Legal Research BenchOverall · All-pass. Higher is better. Test setup not documented as identical

GPT-6 Astra39.423 %

DeepSeek-V4.1-Flash41.346 %

LMArena Code Arena WebDevOverall. Higher is better. No documented setup difference

GPT-6 Astra1,780.244

DeepSeek-V4.1-Flash1,618.513

LMArena Text, Style ControlOverall. Higher is better. No documented setup difference

GPT-6 Astra1,476.716

DeepSeek-V4.1-Flash1,474.579

LMArena Vision, Style ControlOverall. Higher is better. No documented setup difference

GPT-6 Astra1,285.988

DeepSeek-V4.1-Flash1,260.318

MedCodeOverall. Higher is better. Test setup not documented as identical

GPT-6 Astra48.486 %

DeepSeek-V4.1-Flash41.173 %

Key to the marks:better valuebetter, but not conclusivetie

Benchmark scores for GPT-6 Astra and DeepSeek-V4.1-Flash
Benchmark and task	GPT-6 Astra	DeepSeek-V4.1-Flash	Source
BioMysteryBenchOverall Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Harness: Terminus 2, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Harness: Terminus 2, Output limit: 384,000 tokens	Higher measured value79.259 %	67.778 %	Vals AI Source details Data as of: 10/12/2026
Code MigrationOverall Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 384,000 tokens	Higher measured value67.74 %	45.623 %	Vals AI Source details Data as of: 10/12/2026
CyberBench v1.1Overall Test details Test setup not documented as identicalNo direct winnerVersion: 1.1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 384,000 tokens	41.072 %	Higher measured value73.69 %	Vals AI Source details Data as of: 10/12/2026
EMBOverall Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 384,000 tokens	Higher measured value71.7 %	57.208 %	Vals AI Source details Data as of: 10/12/2026
Finance Agent (v2)Overall Test details Test setup not documented as identicalNo direct winnerVersion: 2Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 384,000 tokens	Higher measured value53.54 %	53.481 %	Vals AI Source details Data as of: 10/12/2026
Harvey's Legal Agent BenchmarkOverall · Task fully resolved Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: task resolution rateScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: max, Temperature: 1, Top P: 1, Output limit: 384,000 tokens	5.417 %	Higher measured value6.667 %	Vals AI Source details Data as of: 10/12/2026
IOIOverall Test details Test setup not documented as identicalNo direct winnerVersion: 2Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 384,000 tokens	Higher measured value100 %	40.278 %	Vals AI Source details Data as of: 10/14/2026
Legal Research BenchOverall · All-pass Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: all-pass rateScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: max, Temperature: 1, Top P: 1, Output limit: 384,000 tokens	39.423 %	Higher measured value41.346 %	Vals AI Source details Data as of: 10/12/2026
LMArena Code Arena WebDevOverall Test details No documented difference in the test setupVersion: webdevMetric: WebDev Arena Bradley-Terry ratingScoring: Higher is betterGPT-6 Astra: Snapshot: gpt-6-astra-max, Reasoning: max, Harness: Arena Code Arena WebDev DeepSeek-V4.1-Flash: Snapshot: deepseek-v4.1-flash-max, Reasoning: max, Harness: Arena Code Arena WebDevGPT-6 Astra: Arena rating from 5,746 published pairwise votes.DeepSeek-V4.1-Flash: Arena rating from 3,544 published pairwise votes.	Winner1,780.244 95% confidence interval: ±10.541The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.	1,618.513 95% confidence interval: ±10.935The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.	LMArena Source details Data as of: 10/13/2026
LMArena Text, Style ControlOverall Test details No documented difference in the test setupNo clear leadVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-6 Astra: Snapshot: gpt-6-astra-max, Reasoning: max, Harness: Arena Text, style control DeepSeek-V4.1-Flash: Snapshot: deepseek-v4.1-flash-max, Reasoning: max, Harness: Arena Text, style controlGPT-6 Astra: Arena rating from 11,674 published pairwise votes.DeepSeek-V4.1-Flash: Arena rating from 12,000 published pairwise votes.	No clear lead1,476.716 95% confidence interval: ±6.385The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.	1,474.579 95% confidence interval: ±6.317The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.	LMArena Source details Data as of: 10/13/2026
LMArena Vision, Style ControlOverall Test details No documented difference in the test setupVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-6 Astra: Snapshot: gpt-6-astra-max, Reasoning: max, Harness: Arena Vision, style control DeepSeek-V4.1-Flash: Snapshot: deepseek-v4.1-flash-max, Reasoning: max, Harness: Arena Vision, style controlGPT-6 Astra: Arena rating from 3,590 published pairwise votes.DeepSeek-V4.1-Flash: Arena rating from 3,559 published pairwise votes.	Winner1,285.988 95% confidence interval: ±10.782The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.	1,260.318 95% confidence interval: ±10.899The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.	LMArena Source details Data as of: 10/09/2026
MedCodeOverall Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 384,000 tokens	Higher measured value48.486 %	41.173 %	Vals AI Source details Data as of: 10/13/2026
MedScribeOverall Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 384,000 tokens	Higher measured value87.908 %	85.5 %	Vals AI Source details Data as of: 10/13/2026
MysteryMechanismOverall Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 384,000 tokens	Higher measured value53.153 %	21.171 %	Vals AI Source details Data as of: 10/12/2026
ProgramBenchOverall Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: fully resolved rateScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: max, Temperature: 1, Top P: 1, Output limit: 384,000 tokens	Higher measured value5.5 %	0.5 %	Vals AI Source details Data as of: 10/12/2026
ProofBenchOverall Test details Test setup not documented as identicalNo direct winnerVersion: 1.1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 384,000 tokens	Higher measured value99 %	54 %	Vals AI Source details Data as of: 10/12/2026
SAGEOverall Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 30,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 30,000 tokens	46.367 %	Higher measured value47.88 %	Vals AI Source details Data as of: 10/08/2026
SRE BenchOverall Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: fully solved rateScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: max, Temperature: 1, Top P: 1, Output limit: 384,000 tokens	Higher measured value56.87 %	0.763 %	Vals AI Source details Data as of: 10/14/2026
SWE-bench MultimodalOverall Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: max, Temperature: 1, Top P: 1, Output limit: 384,000 tokens	43.333 %	Higher measured value46.458 %	Vals AI Source details Data as of: 10/13/2026
Tax Agent BenchOverall Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: max, Temperature: 1, Top P: 1, Output limit: 384,000 tokens	20.717 %	Higher measured value26.63 %	Vals AI Source details Data as of: 10/12/2026
Terminal-Bench 2.1Overall Test details Test setup not documented as identicalNo direct winnerVersion: 2.1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 384,000 tokens	Higher measured value87.266 %	74.532 %	Vals AI Source details Data as of: 09/28/2026
Terminal-Bench 4.0Overall Test details Test setup not documented as identicalNo direct winnerVersion: 4.0Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 384,000 tokens	Higher measured value59.596 %	19.697 %	Vals AI Source details Data as of: 10/13/2026
Terminal-Bench ScienceOverall Test details Test setup not documented as identicalNo direct winnerVersion: 0.1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Harness: Mini-SWE-agent, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Harness: Mini-SWE-agent, Output limit: 384,000 tokens	Higher measured value62.857 %	0 %	Vals AI Source details Data as of: 10/14/2026
Vals IndexOverall Test details Test setup not documented as identicalNo direct winnerVersion: 2.1Metric: weighted index scoreScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 384,000 tokens	Higher measured value63.125 %	51.32 %	Vals AI Source details Data as of: 10/13/2026
Vibe Code Bench 1-100Overall Test details Test setup not documented as identicalNo direct winnerVersion: 1.0Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Harness: OpenHands, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Harness: OpenHands, Output limit: 384,000 tokens	Higher measured value27.644 %	16.378 %	Vals AI Source details Data as of: 10/14/2026
Vibe Code Bench v1.1Overall Test details Test setup not documented as identicalNo direct winnerVersion: 1.1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Harness: OpenHands, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Harness: OpenHands, Output limit: 384,000 tokens	Higher measured value89.594 %	84.739 %	Vals AI Source details Data as of: 10/12/2026
BioMysteryBenchHard Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Harness: Terminus 2, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Harness: Terminus 2, Output limit: 384,000 tokens	Higher measured value37.255 %	21.569 %	Vals AI Source details Data as of: 10/12/2026
Code MigrationCLI Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 384,000 tokens	Higher measured value67.463 %	37.497 %	Vals AI Source details Data as of: 10/12/2026
CyberBench v1.1Patch Test details Test setup not documented as identicalNo direct winnerVersion: 1.1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 384,000 tokens	82.143 %	Higher measured value85.714 %	Vals AI Source details Data as of: 10/12/2026
EMBComps Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 384,000 tokens	Higher measured value77.953 %	43.925 %	Vals AI Source details Data as of: 10/12/2026
Finance Agent (v2)Adjustments Test details Test setup not documented as identicalNo direct winnerVersion: 2Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 384,000 tokens	48.318 %	Higher measured value49.241 %	Vals AI Source details Data as of: 10/12/2026
Harvey's Legal Agent BenchmarkAntitrust Competition · Criteria passed Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: criteria pass rateScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: max, Temperature: 1, Top P: 1, Output limit: 384,000 tokens	Higher measured value88.768 %	87.681 %	Vals AI Source details Data as of: 10/12/2026
IOIIOI 2024 Test details Test setup not documented as identicalNo direct winnerVersion: 2Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 384,000 tokens	Higher measured value100 %	42.667 %	Vals AI Source details Data as of: 10/14/2026
Legal Research BenchAdministrative / Regulatory · All-pass Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: all-pass rateScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: max, Temperature: 1, Top P: 1, Output limit: 384,000 tokens	No direct winner53.03 %	No direct winner53.03 %	Vals AI Source details Data as of: 10/12/2026
LMArena Code Arena WebDevImage to webdev Test details No documented difference in the test setupVersion: webdevMetric: WebDev Arena Bradley-Terry ratingScoring: Higher is betterGPT-6 Astra: Snapshot: gpt-6-astra-max, Reasoning: max, Harness: Arena Code Arena WebDev DeepSeek-V4.1-Flash: Snapshot: deepseek-v4.1-flash-max, Reasoning: max, Harness: Arena Code Arena WebDevGPT-6 Astra: Arena rating from 2,739 published pairwise votes.DeepSeek-V4.1-Flash: Arena rating from 766 published pairwise votes.	Winner1,731.395 95% confidence interval: ±13.296The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.	1,608.867 95% confidence interval: ±22.194The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.	LMArena Source details Data as of: 10/09/2026
LMArena Text, Style ControlBusiness, management and finance Test details No documented difference in the test setupNo clear leadVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-6 Astra: Snapshot: gpt-6-astra-max, Reasoning: max, Harness: Arena Text, style control DeepSeek-V4.1-Flash: Snapshot: deepseek-v4.1-flash-max, Reasoning: max, Harness: Arena Text, style controlGPT-6 Astra: Arena rating from 2,061 published pairwise votes.DeepSeek-V4.1-Flash: Arena rating from 2,143 published pairwise votes.	1,472.34 95% confidence interval: ±13.336The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.	No clear lead1,476.217 95% confidence interval: ±12.877The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.	LMArena Source details Data as of: 10/13/2026
LMArena Vision, Style ControlCreative writing vision Test details No documented difference in the test setupNo clear leadVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-6 Astra: Snapshot: gpt-6-astra-max, Reasoning: max, Harness: Arena Vision, style control DeepSeek-V4.1-Flash: Snapshot: deepseek-v4.1-flash-max, Reasoning: max, Harness: Arena Vision, style controlGPT-6 Astra: Arena rating from 367 published pairwise votes.DeepSeek-V4.1-Flash: Arena rating from 299 published pairwise votes.	No clear lead1,289.081 95% confidence interval: ±32.749The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.	1,245.252 95% confidence interval: ±36.257The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.	LMArena Source details Data as of: 10/09/2026
ProgramBenchAlmost Resolved Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: almost-resolved rateScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: max, Temperature: 1, Top P: 1, Output limit: 384,000 tokens	Higher measured value50 %	16.5 %	Vals AI Source details Data as of: 10/12/2026
SAGEAutomata Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 30,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 30,000 tokens	Higher measured value37.302 %	32.936 %	Vals AI Source details Data as of: 10/08/2026
SRE BenchCapability Score Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: capability scoreScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: max, Temperature: 1, Top P: 1, Output limit: 384,000 tokens	Higher measured value63.868 %	14.249 %	Vals AI Source details Data as of: 10/14/2026
SWE-bench Multimodalbpmn-js Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: max, Temperature: 1, Top P: 1, Output limit: 384,000 tokens	Higher measured value62.264 %	60.377 %	Vals AI Source details Data as of: 10/13/2026
Tax Agent BenchControversy & Precedence Analysis Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: max, Temperature: 1, Top P: 1, Output limit: 384,000 tokens	Higher measured value13.333 %	3.333 %	Vals AI Source details Data as of: 10/12/2026
Terminal-Bench 2.1Easy Test details Test setup not documented as identicalNo direct winnerVersion: 2.1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 384,000 tokens	Higher measured value100 %	91.667 %	Vals AI Source details Data as of: 09/28/2026
Terminal-Bench 4.0Hardware Test details Test setup not documented as identicalNo direct winnerVersion: 4.0Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 384,000 tokens	Higher measured value60 %	0 %	Vals AI Source details Data as of: 10/13/2026
Terminal-Bench ScienceEarth Sciences Test details Test setup not documented as identicalNo direct winnerVersion: 0.1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Harness: Mini-SWE-agent, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Harness: Mini-SWE-agent, Output limit: 384,000 tokens	Higher measured value62.5 %	0 %	Vals AI Source details Data as of: 10/14/2026
Vals IndexCode Migration Test details Test setup not documented as identicalNo direct winnerVersion: 2.1Metric: benchmark scoreScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: max, Temperature: 1, Top P: 1, Output limit: 384,000 tokens	Higher measured value68.462 %	46.854 %	Vals AI Source details Data as of: 10/13/2026
BioMysteryBenchHuman-solvable Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Harness: Terminus 2, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Harness: Terminus 2, Output limit: 384,000 tokens	Higher measured value89.041 %	78.539 %	Vals AI Source details Data as of: 10/12/2026
Code MigrationCOBOL Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 384,000 tokens	68.571 %	Higher measured value70 %	Vals AI Source details Data as of: 10/12/2026
CyberBench v1.1PoC Test details Test setup not documented as identicalNo direct winnerVersion: 1.1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 384,000 tokens	0 %	Higher measured value61.667 %	Vals AI Source details Data as of: 10/12/2026
EMBDataroom Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 384,000 tokens	Higher measured value79.319 %	73.863 %	Vals AI Source details Data as of: 10/12/2026
Finance Agent (v2)All-Pass Test details Test setup not documented as identicalNo direct winnerVersion: 2Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 384,000 tokens	40.252 %	Higher measured value41.052 %	Vals AI Source details Data as of: 10/12/2026
Harvey's Legal Agent BenchmarkAntitrust Competition · Task fully resolved Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: task resolution rateScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: max, Temperature: 1, Top P: 1, Output limit: 384,000 tokens	No direct winner0 %	No direct winner0 %	Vals AI Source details Data as of: 10/12/2026
IOIIOI 2025 Test details Test setup not documented as identicalNo direct winnerVersion: 2Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 384,000 tokens	Higher measured value100 %	35.167 %	Vals AI Source details Data as of: 10/14/2026
Legal Research BenchAdministrative / Regulatory · Weighted partial credit Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: weighted partial-credit scoreScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: max, Temperature: 1, Top P: 1, Output limit: 384,000 tokens	87.482 %	Higher measured value89.06 %	Vals AI Source details Data as of: 10/12/2026
LMArena Code Arena WebDevWebdev Test details No documented difference in the test setupVersion: webdevMetric: WebDev Arena Bradley-Terry ratingScoring: Higher is betterGPT-6 Astra: Snapshot: gpt-6-astra-max, Reasoning: max, Harness: Arena Code Arena WebDev DeepSeek-V4.1-Flash: Snapshot: deepseek-v4.1-flash-max, Reasoning: max, Harness: Arena Code Arena WebDevGPT-6 Astra: Arena rating from 5,746 published pairwise votes.DeepSeek-V4.1-Flash: Arena rating from 3,544 published pairwise votes.	Winner1,780.244 95% confidence interval: ±10.541The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.	1,618.513 95% confidence interval: ±10.935The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.	LMArena Source details Data as of: 10/13/2026
LMArena Text, Style ControlChinese Test details No documented difference in the test setupNo clear leadVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-6 Astra: Snapshot: gpt-6-astra-max, Reasoning: max, Harness: Arena Text, style control DeepSeek-V4.1-Flash: Snapshot: deepseek-v4.1-flash-max, Reasoning: max, Harness: Arena Text, style controlGPT-6 Astra: Arena rating from 877 published pairwise votes.DeepSeek-V4.1-Flash: Arena rating from 918 published pairwise votes.	1,514.537 95% confidence interval: ±19.541The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.	No clear lead1,522.52 95% confidence interval: ±19.943The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.	LMArena Source details Data as of: 10/13/2026
LMArena Vision, Style ControlDiagram Test details No documented difference in the test setupNo clear leadVersion: style-control-v1Metric: style-controlled Bradley-Terry ratingScoring: Higher is betterGPT-6 Astra: Snapshot: gpt-6-astra-max, Reasoning: max, Harness: Arena Vision, style control DeepSeek-V4.1-Flash: Snapshot: deepseek-v4.1-flash-max, Reasoning: max, Harness: Arena Vision, style controlGPT-6 Astra: Arena rating from 861 published pairwise votes.DeepSeek-V4.1-Flash: Arena rating from 955 published pairwise votes.	No clear lead1,310.509 95% confidence interval: ±20.739The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.	1,290.828 95% confidence interval: ±19.74The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.	LMArena Source details Data as of: 10/09/2026
ProgramBenchFully Resolved Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: fully resolved rateScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: max, Temperature: 1, Top P: 1, Output limit: 384,000 tokens	Higher measured value5.5 %	0.5 %	Vals AI Source details Data as of: 10/12/2026
SAGECalculus Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 30,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 30,000 tokens	Higher measured value79.938 %	33.333 %	Vals AI Source details Data as of: 10/08/2026
SRE BenchFully Solved Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: fully solved rateScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: max, Temperature: 1, Top P: 1, Output limit: 384,000 tokens	Higher measured value56.87 %	0.763 %	Vals AI Source details Data as of: 10/14/2026
SWE-bench MultimodalCarbon Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: max, Temperature: 1, Top P: 1, Output limit: 384,000 tokens	24.06 %	Higher measured value29.323 %	Vals AI Source details Data as of: 10/13/2026
Tax Agent BenchControversy & Precedence Analysis Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: max, Temperature: 1, Top P: 1, Output limit: 384,000 tokens	Higher measured value58.616 %	52.802 %	Vals AI Source details Data as of: 10/12/2026
Terminal-Bench 2.1Hard Test details Test setup not documented as identicalNo direct winnerVersion: 2.1Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 384,000 tokens	Higher measured value87.778 %	56.667 %	Vals AI Source details Data as of: 09/28/2026
Terminal-Bench 4.0Media Test details Test setup not documented as identicalNo direct winnerVersion: 4.0Metric: accuracyScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4.1-Flash: Reasoning: high, Output limit: 384,000 tokens	Higher measured value66.667 %	8.333 %	Vals AI Source details Data as of: 10/13/2026
Key to the marks:better valuebetter, but not conclusivetie

Our model tests

Our tests

SVG short test

One SVG prompt and fixed output settings. Captures are shown only with 0 reported reasoning tokens or documented lack of reasoning controls. Identical test conditions do not make the models' internal compute identical.

Shared test round: 4ae5d8eeb9a9

GPT-6 Astra

Unavailable in this round

DeepSeek-V4.1-Flash

Earlier tests

Pelican on a bicycle

A community classic for free-form SVG drawing.

GPT-6 Astra

Reasoning (requested): high

Current attempt

Run date: Sep 7, 2026
API provider: openrouter
Requested API model: openai/gpt-6-astra
API model reported in response: openai/gpt-6-astra
Reasoning setting: high
Pinned endpoint: OpenAI | openai/gpt-6-astra-20260903
Temperature: Provider default
Token limit including reasoning: 8,192 tokens
Test protocol: community-visual-tests-v1
Technical checks: Passed, not a quality rating
Response time: 141.7 s

Visually reviewed. This individual test does not establish overall model quality.

DeepSeek-V4.1-Flash

The response reached the token limit. No complete recording is available.

Reasoning (requested): high

Current attempt

Run date: Oct 5, 2026
API provider: openrouter
Requested API model: deepseek/deepseek-v4.1-flash
API model reported in response: deepseek/deepseek-v4.1-flash
Reasoning setting: high
Pinned endpoint: DeepSeek | deepseek/deepseek-v4.1-flash-20260910
Temperature: Provider default
Token limit including reasoning: 8,192 tokens
Test protocol: community-visual-tests-v1
Technical checks: Failed
Response time: 25.8 s

Not yet visually reviewed. A successful recording does not confirm correct physics or full compliance with the prompt.

Earlier attempts (2)

Oct 2, 2026API request failedReasoning setting: high
Oct 1, 2026API request failedReasoning setting: high

Prompt and test conditions

Generate an SVG of a pelican riding a bicycle

Token limit including reasoning: 8,192 tokens

Task origin (Simon Willison)

How to read this comparison

Winning one benchmark is not an overall verdict

Your workload, budget, and required context length matter most. A coding benchmark says little about visual understanding or agent performance.

The standard price table uses direct developer rates. The provider comparison labels OpenRouter endpoints, regions, and special tariffs separately.

Every result links to its measurement source and retrieval date. Labels distinguish vendor reports, official benchmark leaderboards, and independent evaluations. Disputed or archived results are excluded.

More matchups

Compare GPT-6 Astra and DeepSeek-V4.1-Flash with other leading models using the same data and benchmark logic.

View the complete LLM comparison

GPT-6 Astra vs. DeepSeek-V4.1-Flash

Change model selection

General specifications

API pricing compared

Shared benchmarks

Capability profile from matched benchmarks

Our tests

SVG short test

GPT-6 Astra

DeepSeek-V4.1-Flash

Pelican on a bicycle

Winning one benchmark is not an overall verdict

Related LLM comparisons