Head-to-head comparison

GPT-6 Astra vs. DeepSeek-V4-Pro

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

shared result series: 202
distinct benchmark metrics: 16
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-Pro
Attribute	GPT-6 Astra	DeepSeek-V4-Pro
Provider	OpenAI	DeepSeek
Released	Sep 3, 2026	Apr 24, 2026
Availability	Limited access	Active
Model type	Proprietary	Open Weights
Parameters	Not published	1.6T, 49B active
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, Context

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-Pro
Price type	GPT-6 Astra	DeepSeek-V4-Pro
API input	$10 ≤272K / $20 >272K	$0.66 (Off-Peak)
API output	$50 ≤272K / $75 >272K	$1.98 (Off-Peak)
Cache read	$1 ≤272K / $2 >272K	$0.022 (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	09/22/2026 Source
DeepSeek-V4-Pro: peak pricing is $1.32 input / $3.96 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-Pro

API inputUSD per 1M tokens, base tier

GPT-6 Astra$10

DeepSeek-V4-Pro$0.66

API outputUSD per 1M tokens, base tier

GPT-6 Astra$50

DeepSeek-V4-Pro$1.98

Cache readUSD per 1M tokens, base tier

GPT-6 Astra$1

DeepSeek-V4-Pro$0.022

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-Pro
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	Source10/19/2026
	Azure	Via OpenRouterazure	$10	$50	Source10/19/2026
	Azure	Via OpenRouterazure/us	$11	$55	Source10/19/2026
	OpenAI	Via OpenRouteropenai	$10	$50	Source10/19/2026
	OpenAI	Via OpenRouteropenai/fast	$20	$100	Source10/19/2026
	OpenAI	Via OpenRouteropenai/flex	Lowest listed price: $5	Lowest listed price: $25	Source10/19/2026
	OpenAI	Via OpenRouteropenai/ultrafast	$60	$300	Source10/19/2026
DeepSeek-V4-Pro	DeepSeek	Direct from developerStandard API	$0.66	$1.98	Source09/22/2026
	Alibaba	Via OpenRouteralibaba	$0.5808	$1.7424	Source10/19/2026
	Alibaba	Via OpenRouteralibaba/us-east-1	$0.66	$1.98	Source10/19/2026
	AtlasCloud	Via OpenRouteratlas-cloud/fp8	$1.32	$3.96	Source10/19/2026
	Baidu	Via OpenRouterbaidu/fp8	$0.1333	Lowest listed price: $0.4	Source10/19/2026
	Cloudflare	Via OpenRoutercloudflare	$1.32	$3.96	Source10/19/2026
	CoreWeave	Via OpenRoutercoreweave/fp8	$1.31	$3.96	Source10/19/2026
	DeepInfra	Via OpenRouterdeepinfra/fp8	$1.3	$2.6	Source10/19/2026
	DeepSeek	Via OpenRouterdeepseek	$0.66	$1.98	Source10/19/2026
	DekaLLM	Via OpenRouterdekallm	$0.88	$3.8	Source10/19/2026
	DigitalOcean	Via OpenRouterdigitalocean	$1.32	$3.96	Source10/19/2026
	GMICloud	Via OpenRoutergmicloud/fp8	$1.056	$3.168	Source10/19/2026
	Ionstream	Via OpenRouterionstream	$0.37	$2.93	Source10/19/2026
	Novita	Via OpenRouternovita/fp8	$0.99	$2.97	Source10/19/2026
	Parasail	Via OpenRouterparasail/fp8	$1.32	$3.96	Source10/19/2026
	Phala	Via OpenRouterphala	$0.957	$2.8776	Source10/19/2026
	Relace	Via OpenRouterrelace/fp4	Lowest listed price: $0.12	$8	Source10/19/2026
	Sail Research	Via OpenRoutersail-research/fp4	$0.896	$0.582	Source10/19/2026
	Sail Research	Via OpenRoutersail-research/us	$0.896	$0.582	Source10/19/2026
	SiliconFlow	Via OpenRoutersiliconflow/fp8	$1.32	$3.96	Source10/19/2026
	StreamLake	Via OpenRouterstreamlake	$0.66	$1.98	Source10/19/2026
	Together	Via OpenRoutertogether	$1.32	$3.96	Source10/19/2026
	Venice	Via OpenRoutervenice	$1.65	$4.95	Source10/19/2026
	Wafer	Via OpenRouterwafer/mxfp4	$0.25	$5	Source10/19/2026

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.

Not enough directly matched categories for a reliable radar

Only 2 of at least 5 categories have comparable, direct model measurements. The benchmark tables below show the available evidence without filling missing axes.

GPT-6 AstraDeepSeek-V4-Pro

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

GPT-6 Astra67.74 %

DeepSeek-V4-Pro26.201 %

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

GPT-6 Astra71.7 %

DeepSeek-V4-Pro51.623 %

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

GPT-6 Astra53.54 %

DeepSeek-V4-Pro44.083 %

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

GPT-6 Astra5.417 %

DeepSeek-V4-Pro3.75 %

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

GPT-6 Astra39.423 %

DeepSeek-V4-Pro23.077 %

LMArena Code Arena WebDevOverall. Higher is better. Test setup not documented as identical

GPT-6 Astra1,780.244

DeepSeek-V4-Pro1,446.815

LMArena Text, Style ControlOverall. Higher is better. Test setup not documented as identical

GPT-6 Astra1,476.716

DeepSeek-V4-Pro1,457.848

MedCodeOverall. Higher is better. No documented setup difference

GPT-6 Astra48.486 %

DeepSeek-V4-Pro40.455 %

MedScribeOverall. Higher is better. No documented setup difference

GPT-6 Astra87.908 %

DeepSeek-V4-Pro75.144 %

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

GPT-6 Astra5.5 %

DeepSeek-V4-Pro0 %

ProofBenchOverall. Higher is better. No documented setup difference

GPT-6 Astra99 %

DeepSeek-V4-Pro16 %

Tax Agent BenchOverall. Higher is better. Test setup not documented as identical

GPT-6 Astra20.717 %

DeepSeek-V4-Pro25.17 %

Key to the marks:better valuebetter, but not conclusivetie

Benchmark scores for GPT-6 Astra and DeepSeek-V4-Pro
Benchmark and task	GPT-6 Astra	DeepSeek-V4-Pro	Source
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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value67.74 %	26.201 %	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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value71.7 %	51.623 %	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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value53.54 %	44.083 %	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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value5.417 %	3.75 %	Vals AI Source details Data as of: 10/12/2026
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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value39.423 %	23.077 %	Vals AI Source details Data as of: 10/12/2026
LMArena Code Arena WebDevOverall Test details Test setup not documented as identicalNo direct winnerVersion: 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-Pro: Snapshot: deepseek-v4-pro, Harness: Arena Code Arena WebDevGPT-6 Astra: Arena rating from 5,746 published pairwise votes.DeepSeek-V4-Pro: Arena rating from 14,023 published pairwise votes.	Higher measured value1,780.244 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.	1,446.815 95% confidence interval: ±6.319The 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 Test setup not documented as identicalNo direct winnerVersion: 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-Pro: Snapshot: deepseek-v4-pro, Harness: Arena Text, style controlGPT-6 Astra: Arena rating from 11,674 published pairwise votes.DeepSeek-V4-Pro: Arena rating from 57,614 published pairwise votes.	Higher measured value1,476.716 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.	1,457.848 95% confidence interval: ±3.899The 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: 10/13/2026
MedCodeOverall Test details No documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is betterSettings: Reasoning: max, Output limit: 128,000 tokens	Winner48.486 %	40.455 %	Vals AI Source details Data as of: 10/13/2026
MedScribeOverall Test details No documented difference in the test setupVersion: 1Metric: accuracyScoring: Higher is betterSettings: Reasoning: max, Output limit: 128,000 tokens	Winner87.908 %	75.144 %	Vals AI Source details Data as of: 10/13/2026
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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value5.5 %	0 %	Vals AI Source details Data as of: 10/12/2026
ProofBenchOverall Test details No documented difference in the test setupVersion: 1.1Metric: accuracyScoring: Higher is betterSettings: Reasoning: max, Output limit: 128,000 tokens	Winner99 %	16 %	Vals AI Source details Data as of: 10/12/2026
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-Pro: Reasoning: max, Output limit: 384,000 tokens	20.717 %	Higher measured value25.17 %	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-Pro: Output limit: 128,000 tokens	Higher measured value87.266 %	50.187 %	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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value59.596 %	11.111 %	Vals AI Source details Data as of: 10/13/2026
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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value63.125 %	38.63 %	Vals AI Source details Data as of: 10/13/2026
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-Pro: Reasoning: max, Harness: OpenHands, Output limit: 256,000 tokens	Higher measured value89.594 %	49.931 %	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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value67.463 %	15.47 %	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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value77.953 %	57.434 %	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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value48.318 %	41.586 %	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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value88.768 %	83.696 %	Vals AI Source details Data as of: 10/12/2026
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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value53.03 %	28.788 %	Vals AI Source details Data as of: 10/12/2026
LMArena Code Arena WebDevOverall Test details Test setup not documented as identicalNo direct winnerVersion: 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-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena Code Arena WebDevGPT-6 Astra: Arena rating from 5,746 published pairwise votes.DeepSeek-V4-Pro: Arena rating from 13,199 published pairwise votes.	Higher measured value1,780.244 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.	1,463.849 95% confidence interval: ±6.309The 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 Test setup not documented as identicalNo direct winnerVersion: 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-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena Text, style controlGPT-6 Astra: Arena rating from 11,674 published pairwise votes.DeepSeek-V4-Pro: Arena rating from 54,929 published pairwise votes.	Higher measured value1,476.716 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.	1,454.814 95% confidence interval: ±3.964The 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
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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value50 %	0 %	Vals AI Source details Data as of: 10/12/2026
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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value13.333 %	12.333 %	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-Pro: Output limit: 128,000 tokens	Higher measured value100 %	66.667 %	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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value60 %	0 %	Vals AI Source details Data as of: 10/13/2026
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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value68.462 %	26.212 %	Vals AI Source details Data as of: 10/13/2026
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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value68.571 %	58.393 %	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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value79.319 %	62.572 %	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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value40.252 %	31.448 %	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-Pro: Reasoning: max, Output limit: 384,000 tokens	No direct winner0 %	No direct winner0 %	Vals AI Source details Data as of: 10/12/2026
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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value87.482 %	83.372 %	Vals AI Source details Data as of: 10/12/2026
LMArena Code Arena WebDevWebdev Test details Test setup not documented as identicalNo direct winnerVersion: 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-Pro: Snapshot: deepseek-v4-pro, Harness: Arena Code Arena WebDevGPT-6 Astra: Arena rating from 5,746 published pairwise votes.DeepSeek-V4-Pro: Arena rating from 14,023 published pairwise votes.	Higher measured value1,780.244 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.	1,446.815 95% confidence interval: ±6.319The 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 ControlBusiness, management and finance Test details Test setup not documented as identicalNo direct winnerVersion: 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-Pro: Snapshot: deepseek-v4-pro, Harness: Arena Text, style controlGPT-6 Astra: Arena rating from 2,061 published pairwise votes.DeepSeek-V4-Pro: Arena rating from 11,383 published pairwise votes.	Higher measured value1,472.34 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.	1,459.786 95% confidence interval: ±6.716The 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
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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value5.5 %	0 %	Vals AI Source details Data as of: 10/12/2026
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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value58.616 %	54.843 %	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-Pro: Output limit: 128,000 tokens	Higher measured value87.778 %	31.111 %	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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value66.667 %	0 %	Vals AI Source details Data as of: 10/13/2026
Vals IndexEMB 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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value71.7 %	51.623 %	Vals AI Source details Data as of: 10/13/2026
Code MigrationCode Quality Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: mean reviewer code-quality ratingScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value4.469 / 5	3.401 / 5	Vals AI Source details Data as of: 10/12/2026
EMBDCF 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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value64.483 %	44.435 %	Vals AI Source details Data as of: 10/12/2026
Finance Agent (v2)Comparables 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-Pro: Reasoning: max, Output limit: 384,000 tokens	37.758 %	Higher measured value39.632 %	Vals AI Source details Data as of: 10/12/2026
Harvey's Legal Agent BenchmarkArbitration International Dispute Resolution · 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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value86.022 %	83.513 %	Vals AI Source details Data as of: 10/12/2026
Legal Research BenchBusiness & Commercial · 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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value32.558 %	23.256 %	Vals AI Source details Data as of: 10/12/2026
LMArena Code Arena WebDevWebdev Test details Test setup not documented as identicalNo direct winnerVersion: 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-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena Code Arena WebDevGPT-6 Astra: Arena rating from 5,746 published pairwise votes.DeepSeek-V4-Pro: Arena rating from 13,199 published pairwise votes.	Higher measured value1,780.244 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.	1,463.849 95% confidence interval: ±6.309The 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 ControlBusiness, management and finance Test details Test setup not documented as identicalNo direct winnerVersion: 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-Pro: Snapshot: deepseek-v4-pro-high-preview, Reasoning: high, Harness: Arena Text, style controlGPT-6 Astra: Arena rating from 2,061 published pairwise votes.DeepSeek-V4-Pro: Arena rating from 11,045 published pairwise votes.	Higher measured value1,472.34 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.	1,448.007 95% confidence interval: ±6.811The 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
ProgramBenchRaw Pass Rate Test details Test setup not documented as identicalNo direct winnerVersion: 1Metric: raw test pass rateScoring: Higher is betterGPT-6 Astra: Reasoning: max, Output limit: 128,000 tokens DeepSeek-V4-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value85.417 %	47.778 %	Vals AI Source details Data as of: 10/12/2026
Tax Agent BenchCurrent & Temporal 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-Pro: Reasoning: max, Output limit: 384,000 tokens	29.944 %	Higher measured value36.595 %	Vals AI Source details Data as of: 10/12/2026
Terminal-Bench 2.1Medium 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-Pro: Output limit: 128,000 tokens	Higher measured value86.061 %	59.394 %	Vals AI Source details Data as of: 09/28/2026
Terminal-Bench 4.0ML 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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value66.667 %	3.03 %	Vals AI Source details Data as of: 10/13/2026
Vals IndexFinance Agent v2 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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value53.54 %	44.083 %	Vals AI Source details Data as of: 10/13/2026
EMBLBO 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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value61.825 %	43.938 %	Vals AI Source details Data as of: 10/12/2026
Finance Agent (v2)Disclosure Analysis 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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value61.684 %	52.649 %	Vals AI Source details Data as of: 10/12/2026
Harvey's Legal Agent BenchmarkArbitration International Dispute Resolution · 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-Pro: Reasoning: max, Output limit: 384,000 tokens	No direct winner0 %	No direct winner0 %	Vals AI Source details Data as of: 10/12/2026
Legal Research BenchBusiness & Commercial · 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-Pro: Reasoning: max, Output limit: 384,000 tokens	83.821 %	Higher measured value86.491 %	Vals AI Source details Data as of: 10/12/2026
LMArena Code Arena WebDevWebdev-html Test details Test setup not documented as identicalNo direct winnerVersion: 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-Pro: Snapshot: deepseek-v4-pro, Harness: Arena Code Arena WebDevGPT-6 Astra: Arena rating from 1,129 published pairwise votes.DeepSeek-V4-Pro: Arena rating from 1,502 published pairwise votes.	Higher measured value1,780.13 95% confidence interval: ±22.887The Arena score measures human preference rather than a fixed capability test. The confidence interval, Arena category, and model configuration are part of the result.	1,434.123 95% confidence interval: ±15.907The 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 Test setup not documented as identicalNo direct winnerVersion: 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-Pro: Snapshot: deepseek-v4-pro, Harness: Arena Text, style controlGPT-6 Astra: Arena rating from 877 published pairwise votes.DeepSeek-V4-Pro: Arena rating from 3,378 published pairwise votes.	Higher measured value1,514.537 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.	1,488.803 95% confidence interval: ±10.784The 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
Tax Agent BenchCurrent & Temporal 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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value66.224 %	65.741 %	Vals AI Source details Data as of: 10/12/2026
Terminal-Bench 4.0Operations 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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value48.148 %	7.407 %	Vals AI Source details Data as of: 10/13/2026
Vals IndexHLAB 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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value5.417 %	3.75 %	Vals AI Source details Data as of: 10/13/2026
EMBM&A 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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value83.066 %	48.156 %	Vals AI Source details Data as of: 10/12/2026
Finance Agent (v2)Earnings Analysis 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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value69.474 %	59.84 %	Vals AI Source details Data as of: 10/12/2026
Harvey's Legal Agent BenchmarkBanking Finance · 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-Pro: Reasoning: max, Output limit: 384,000 tokens	Higher measured value88.104 %	86.245 %	Vals AI Source details Data as of: 10/12/2026
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-Pro

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-Pro

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-pro-0813
API model reported in response: deepseek/deepseek-v4-pro-0813
Reasoning setting: high
Pinned endpoint: DeepSeek | deepseek/deepseek-v4-pro-20260813
Temperature: Provider default
Token limit including reasoning: 8,192 tokens
Test protocol: community-visual-tests-v1
Technical checks: Failed
Response time: 67.6 s

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

Earlier attempts (3)

Oct 2, 2026API request failedReasoning setting: high
Oct 1, 2026API request failedReasoning setting: high
Sep 7, 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-Pro with other leading models using the same data and benchmark logic.

View the complete LLM comparison

GPT-6 Astra vs. DeepSeek-V4-Pro

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-Pro

Pelican on a bicycle

Winning one benchmark is not an overall verdict

Related LLM comparisons