Skip to main content

ByteDanceLanguage modelDeprecated

Seed 1.6 Flash

Seed 1.6 Flash has no overall index value yet because there are too few independent benchmark results.

Intelligencenot ranked yet

n/aOverall indexhigher is better

Speednot measured

n/ano measurement yet

PriceRank 4 of 95

Input $0.075 · Output $0.30

$0.13per 1M tokenslower is better

ContextRank 76 of 102

256Ktokens at once≈ 380 A4 pages

PopularityRank 634 of 875

9out of 100same as last month

Summary

It is cheap: $0.13 per 1M tokens (input and output blended); a typical model costs $1.56.

Seed 1.6 Flash is ByteDance’s compact API variant of its multimodal deep-thinking model. It accepts text, images, and video, supports up to 256K tokens of context, and can answer directly with thinking disabled. The parameter count is unpublished, and no weights are provided.

BytePlus documents the service as a controllable API for production agent and document work. There is no checkpoint for local research because ByteDance provides only the provider-bound serving form.

See for yourself: one prompt, one attempt

SVG short test

One SVG task with the same fixed settings for every model, without hidden thinking.

Seed 1.6 Flash

Seed 1.6 Flash: SVG short test

Prompt

Return only one standalone SVG of a pelican riding a bicycle. Use viewBox="0 0 800 600", simple shapes, two clearly visible wheels and a recognizable pelican with a long beak. At most 30 SVG elements and 1800 characters. No Markdown, explanations, comments, scripts, animation, CSS or external resources.

Oct 14, 2026 · $0.0003

Test details
Test run
Oct 14, 2026
Provider
seed/fp8
API model
bytedance-seed/seed-1.6-flash
Output
1,054 tokens
Reasoning
API reports 0 reasoning tokens
Cost of the attempt
$0.0003
Test round
43b8c5c46830
Earlier tests
Seed 1.6 Flash: Pelican on a bicycle
A community classic for free-form SVG drawing.

Pelican on a bicycle

Prompt

Generate an SVG of a pelican riding a bicycle

Oct 1, 2026 · 28 s · $0.0016

Test details
Run date
Oct 1, 2026
API provider
openrouter
Requested API model
bytedance-seed/seed-1.6-flash
Reasoning setting
high
Token limit including reasoning
8,192 tokens
Output tokens
5,265 tokens
Cost of the attempt
$0.0016
Response time
28 s
API model reported in response
bytedance-seed/seed-1.6-flash
Pinned endpoint
Seed | bytedance-seed/seed-1.6-flash-20250625
Temperature
Provider default
Test protocol
community-visual-tests-v1
Technical checks
Passed, not a quality rating

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

Task origin: Simon Willison

All benchmark results

Every published result with its source, test setup and comparison models.

Show all 21 published benchmark results

Page 1 of 2

Measurements without matching peer values

These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.

Memetic Landscape aggregate accuracy (Show measurement, test conditions, and source)
Source value
61
Score
61
Metric
Acc
Unit
%
Benchmark version
Paper Table 3; aggregate over MMLU-Redux, MATH-500, and SimpleQA
Category
general
Direction
Higher is better
Source type
independent-evaluator
Evaluator
A Memetic Landscape for Scrutinizing LLMs, OpenReview Table 3
Status
active
Retrieved at
2026-10-15
Methodology
Independent evaluation. Exact paper row doubao-seed-1-6-flash-250715 (IR). Acc 61.0 as the aggregate over MMLU-Redux, MATH-500, and SimpleQA. Table 3 reports aggregate accuracy over MMLU-Redux, MATH-500, and SimpleQA. The paper does not disclose task date, prompt/harness, decoding, quantization, or sample count; IR is retained literally and not expanded.
Context
This result is for the dated 250715 checkpoint, not every Seed 1.6 Flash API alias. The paper's aggregate and literal setting labels are retained as reported.
Source A Memetic Landscape for Scrutinizing LLMs, OpenReview Table 3
Memetic Landscape aggregate accuracy (Show measurement, test conditions, and source)
Source value
56.7
Score
56.7
Metric
Acc
Unit
%
Benchmark version
Paper Table 3; aggregate over MMLU-Redux, MATH-500, and SimpleQA
Category
general
Direction
Higher is better
Source type
independent-evaluator
Evaluator
A Memetic Landscape for Scrutinizing LLMs, OpenReview Table 3
Status
active
Retrieved at
2026-10-15
Methodology
Independent evaluation. Exact paper row doubao-seed-1-6-flash-250715 (CoT). Acc 56.7 as the aggregate over MMLU-Redux, MATH-500, and SimpleQA. Table 3 reports aggregate accuracy over MMLU-Redux, MATH-500, and SimpleQA. The paper does not disclose task date, prompt/harness, decoding, quantization, or sample count; CoT is retained literally and not expanded.
Context
This result is for the dated 250715 checkpoint, not every Seed 1.6 Flash API alias. The paper's aggregate and literal setting labels are retained as reported.
Source A Memetic Landscape for Scrutinizing LLMs, OpenReview Table 3
AEGIS FSD accuracy (Show measurement, test conditions, and source)
Source value
29.26
Score
29.26
Metric
single-choice classification accuracy (ACC)
Unit
%
Benchmark version
AEGIS Table 2; ACL 2026 Long Papers
Category
multimodal
Direction
Higher is better
Source type
independent-evaluator
Evaluator
AEGIS, ACL 2026, Table 2
Status
active
Retrieved at
2026-10-16
Methodology
Independent evaluation. Exact Table 2 model label Doubao-Seed-1.6-flash, mapped only to the matching seed-1-6-flash API-alias profile. Appendix B.1 says the Doubao series used the Volcano Engine API; Appendix B.2 lists the corresponding official console route. AEGIS evaluated image-forensics tasks on high-resolution PNGs with minimal task-specific prompts. The paper does not identify an immutable serving revision or evaluation request date. Per-task denominators, temperature, and other decoding settings are not reported; the 8,210-sample dataset total is not treated as this metric's N.
Context
This is an AEGIS academic-image-forensics task result, not a general model-quality score. The paper names an API alias without a request date or immutable backend revision. It is not attributed to the separately reported doubao-seed-1-6-flash-250715 checkpoint.
Source AEGIS, ACL 2026, Table 2
AEGIS FSD macro-F1 (Show measurement, test conditions, and source)
Source value
23.77
Score
23.77
Metric
class-balanced macro-F1 (M-F1)
Unit
%
Benchmark version
AEGIS Table 2; ACL 2026 Long Papers
Category
multimodal
Direction
Higher is better
Source type
independent-evaluator
Evaluator
AEGIS, ACL 2026, Table 2
Status
active
Retrieved at
2026-10-16
Methodology
Independent evaluation. Exact Table 2 model label Doubao-Seed-1.6-flash, mapped only to the matching seed-1-6-flash API-alias profile. Appendix B.1 says the Doubao series used the Volcano Engine API; Appendix B.2 lists the corresponding official console route. AEGIS evaluated image-forensics tasks on high-resolution PNGs with minimal task-specific prompts. The paper does not identify an immutable serving revision or evaluation request date. Per-task denominators, temperature, and other decoding settings are not reported; the 8,210-sample dataset total is not treated as this metric's N.
Context
This is an AEGIS academic-image-forensics task result, not a general model-quality score. The paper names an API alias without a request date or immutable backend revision. It is not attributed to the separately reported doubao-seed-1-6-flash-250715 checkpoint.
Source AEGIS, ACL 2026, Table 2
AEGIS TAR accuracy (Show measurement, test conditions, and source)
Source value
60.67
Score
60.67
Metric
single-choice classification accuracy (ACC)
Unit
%
Benchmark version
AEGIS Table 2; ACL 2026 Long Papers
Category
multimodal
Direction
Higher is better
Source type
independent-evaluator
Evaluator
AEGIS, ACL 2026, Table 2
Status
active
Retrieved at
2026-10-16
Methodology
Independent evaluation. Exact Table 2 model label Doubao-Seed-1.6-flash, mapped only to the matching seed-1-6-flash API-alias profile. Appendix B.1 says the Doubao series used the Volcano Engine API; Appendix B.2 lists the corresponding official console route. AEGIS evaluated image-forensics tasks on high-resolution PNGs with minimal task-specific prompts. The paper does not identify an immutable serving revision or evaluation request date. Per-task denominators, temperature, and other decoding settings are not reported; the 8,210-sample dataset total is not treated as this metric's N.
Context
This is an AEGIS academic-image-forensics task result, not a general model-quality score. The paper names an API alias without a request date or immutable backend revision. It is not attributed to the separately reported doubao-seed-1-6-flash-250715 checkpoint.
Source AEGIS, ACL 2026, Table 2
AEGIS TAR macro-F1 (Show measurement, test conditions, and source)
Source value
60.39
Score
60.39
Metric
class-balanced macro-F1 (M-F1)
Unit
%
Benchmark version
AEGIS Table 2; ACL 2026 Long Papers
Category
multimodal
Direction
Higher is better
Source type
independent-evaluator
Evaluator
AEGIS, ACL 2026, Table 2
Status
active
Retrieved at
2026-10-16
Methodology
Independent evaluation. Exact Table 2 model label Doubao-Seed-1.6-flash, mapped only to the matching seed-1-6-flash API-alias profile. Appendix B.1 says the Doubao series used the Volcano Engine API; Appendix B.2 lists the corresponding official console route. AEGIS evaluated image-forensics tasks on high-resolution PNGs with minimal task-specific prompts. The paper does not identify an immutable serving revision or evaluation request date. Per-task denominators, temperature, and other decoding settings are not reported; the 8,210-sample dataset total is not treated as this metric's N.
Context
This is an AEGIS academic-image-forensics task result, not a general model-quality score. The paper names an API alias without a request date or immutable backend revision. It is not attributed to the separately reported doubao-seed-1-6-flash-250715 checkpoint.
Source AEGIS, ACL 2026, Table 2
AEGIS MC accuracy (Show measurement, test conditions, and source)
Source value
35.89
Score
35.89
Metric
single-choice classification accuracy (ACC)
Unit
%
Benchmark version
AEGIS Table 2; ACL 2026 Long Papers
Category
multimodal
Direction
Higher is better
Source type
independent-evaluator
Evaluator
AEGIS, ACL 2026, Table 2
Status
active
Retrieved at
2026-10-16
Methodology
Independent evaluation. Exact Table 2 model label Doubao-Seed-1.6-flash, mapped only to the matching seed-1-6-flash API-alias profile. Appendix B.1 says the Doubao series used the Volcano Engine API; Appendix B.2 lists the corresponding official console route. AEGIS evaluated image-forensics tasks on high-resolution PNGs with minimal task-specific prompts. The paper does not identify an immutable serving revision or evaluation request date. Per-task denominators, temperature, and other decoding settings are not reported; the 8,210-sample dataset total is not treated as this metric's N.
Context
This is an AEGIS academic-image-forensics task result, not a general model-quality score. The paper names an API alias without a request date or immutable backend revision. It is not attributed to the separately reported doubao-seed-1-6-flash-250715 checkpoint.
Source AEGIS, ACL 2026, Table 2
AEGIS MC macro-F1 (Show measurement, test conditions, and source)
Source value
30.96
Score
30.96
Metric
class-balanced macro-F1 (M-F1)
Unit
%
Benchmark version
AEGIS Table 2; ACL 2026 Long Papers
Category
multimodal
Direction
Higher is better
Source type
independent-evaluator
Evaluator
AEGIS, ACL 2026, Table 2
Status
active
Retrieved at
2026-10-16
Methodology
Independent evaluation. Exact Table 2 model label Doubao-Seed-1.6-flash, mapped only to the matching seed-1-6-flash API-alias profile. Appendix B.1 says the Doubao series used the Volcano Engine API; Appendix B.2 lists the corresponding official console route. AEGIS evaluated image-forensics tasks on high-resolution PNGs with minimal task-specific prompts. The paper does not identify an immutable serving revision or evaluation request date. Per-task denominators, temperature, and other decoding settings are not reported; the 8,210-sample dataset total is not treated as this metric's N.
Context
This is an AEGIS academic-image-forensics task result, not a general model-quality score. The paper names an API alias without a request date or immutable backend revision. It is not attributed to the separately reported doubao-seed-1-6-flash-250715 checkpoint.
Source AEGIS, ACL 2026, Table 2
AEGIS TP region-level Correct Localization Accuracy (Show measurement, test conditions, and source)
Source value
52.36
Score
52.36
Metric
sample-level region localization accuracy (CLA); coverage threshold τ=0.5
Unit
%
Benchmark version
AEGIS Table 2; ACL 2026 Long Papers
Category
multimodal
Direction
Higher is better
Source type
independent-evaluator
Evaluator
AEGIS, ACL 2026, Table 2
Status
active
Retrieved at
2026-10-16
Methodology
Independent evaluation. Exact Table 2 model label Doubao-Seed-1.6-flash, mapped only to the matching seed-1-6-flash API-alias profile. Appendix B.1 says the Doubao series used the Volcano Engine API; Appendix B.2 lists the corresponding official console route. AEGIS evaluated image-forensics tasks on high-resolution PNGs with minimal task-specific prompts. For MLLMs, TP localization uses region-level bounding boxes. CLA counts a sample correct when each ground-truth tampered region is at least 50% covered. The paper does not identify an immutable serving revision, evaluation request date, or per-task denominator.
Context
This is an AEGIS academic-image-forensics localization result, not a general model-quality score. The paper names an API alias without a request date or immutable backend revision. It is not attributed to the separately reported doubao-seed-1-6-flash-250715 checkpoint.
Source AEGIS, ACL 2026, Table 2
AEGIS TP Over Localization Rate (Show measurement, test conditions, and source)
Source value
1.78
Score
1.78
Metric
percentage of samples with over-localized regions (OLR); lower is better
Unit
%
Benchmark version
AEGIS Table 2; ACL 2026 Long Papers
Category
multimodal
Direction
Lower is better
Source type
independent-evaluator
Evaluator
AEGIS, ACL 2026, Table 2
Status
active
Retrieved at
2026-10-16
Methodology
Independent evaluation. Exact Table 2 model label Doubao-Seed-1.6-flash, mapped only to the matching seed-1-6-flash API-alias profile. Appendix B.1 says the Doubao series used the Volcano Engine API; Appendix B.2 lists the corresponding official console route. AEGIS evaluated image-forensics tasks on high-resolution PNGs with minimal task-specific prompts. For MLLMs, OLR is the percentage of evaluated samples with excessive predicted regions under a 0.5 over-localization tolerance. Lower values mean fewer samples met the paper's over-localization condition. The paper does not identify an immutable serving revision, evaluation request date, or per-task denominator.
Context
This is an AEGIS pipeline-specific over-localization rate, where lower is better, not a general model-quality score. The paper names an API alias without a request date or immutable backend revision. It is not attributed to the separately reported doubao-seed-1-6-flash-250715 checkpoint.
Source AEGIS, ACL 2026, Table 2
AEGIS Normalized Forensic Index (Show measurement, test conditions, and source)
Source value
36.03
Score
36.03
Metric
paper-defined 0-100 composite of FSD-F1, TAR-F1, MC-F1, CLA, with OLR penalty
Unit
points
Benchmark version
AEGIS Table 2; ACL 2026 Long Papers; default OLR exponent γ=0.5
Category
multimodal
Direction
Higher is better
Source type
independent-evaluator
Evaluator
AEGIS, ACL 2026, Table 2
Status
active
Retrieved at
2026-10-16
Methodology
Independent evaluation. Exact Table 2 model label Doubao-Seed-1.6-flash, mapped only to the matching seed-1-6-flash API-alias profile. Appendix B.1 says the Doubao series used the Volcano Engine API; Appendix B.2 lists the corresponding official console route. AEGIS NFI is a paper-defined 0-100 composite: the harmonic mean of FSD-F1, TAR-F1, MC-F1, and CLA, multiplied by an over-localization penalty (1-OLR)^γ with default γ=0.5. The paper says higher NFI indicates stronger balanced performance on its image-forensics tasks. This composite is not a general model-quality index. The paper does not identify an immutable serving revision or evaluation request date.
Context
This is the authors' AEGIS-specific composite, not a broad model-quality score or an index comparable across unrelated benchmarks. The paper names an API alias without a request date or immutable backend revision. It is not attributed to the separately reported doubao-seed-1-6-flash-250715 checkpoint.
Source AEGIS, ACL 2026, Table 2
The Aggregate published composite rating (Show measurement, test conditions, and source)
Source value
1602
Score
1602
Metric
Aggregate Rating (publisher composite)
Unit
rating
Benchmark version
snapshot 2026-10-18
Direction
Higher is better
Source type
independent-evaluator
Evaluator
The Aggregate, ByteDance model ratings
Status
active
Retrieved at
2026-10-19
Methodology
Independent evaluation. The Aggregate's independently published model-level composite rating, combining public leaderboard and benchmark-report scores through hierarchical continuous-response IRT. Its scale is Elo-like around 1500, not a raw score on one task. The exact labels on the company table are used without alias transfer.
Context
The company page labels its table snapshot 2026-10-18, while the method-page footer says 2026-10-19. The table displays a row-specific ± value; the method page describes conditional standard errors, not 95% confidence intervals. No sample count is given. This is a daily changing publisher composite, not a raw task result.
Source The Aggregate, ByteDance model ratings

Page 1 of 2

Popularity compared

How often the model is searched for, used and mentioned, compared with every model of its category.

Popularity index

The 100 most popular models of this category, 0 to 100, higher is more popular

20 of 101 models

Gemini 4 Argon92
GPT-6 Astra78
Claude Haiku 5.577
Qwen3.5 4B76
GPT-573
GPT-6 Sol72
Qwen3.5 0.8B72
Qwen3.5 2B71
Qwen 3.8 27B70
Claude Opus 5.570
DeepSeek V4.1 Flash 101768
Qwen 3.8 Max 090268
Kimi K368
DeepSeek-V4-Flash65
GPT-5.6 Terra64
Claude Sonnet 563
DiffusionGemma 26B A4B IT63
Claude Opus 562
Qwen 3.8 Flash62
Seed 1.6 Flash9

As of November 2026 · Source: gradually.ai popularity index

Over time

Weekly values from October 18, 2026 to November 5, 2026

Seed 1.6 FlashJamba MiniClaude Mythos 5.1Kimi K2 Thinking TurboGPT-6.1 Terra

All specifications

Input and output

Context window
256,000 tokens
Source: Context window
Input
Text, Image, Video (retrieved 2026-09-22)
Source: Input
Output
Text (retrieved 2026-09-22)
Source: Output
Capabilities
Function calling, context caching and structured output (Beta). Also processes images and videos. (retrieved 2026-09-22)
Source: Capabilities

Watermarking and provenance

AI watermarking refers to invisible signals embedded in the output. Visible marks and provenance metadata are listed separately. Unknown means there is insufficient evidence for that particular field.

AI watermarking
Unknown
no source
Watermarking technique
Not known
no source

Licence and access

Weights
Proprietary
Source: Weights
Access
API
Source: Access
Release
Jun 15, 2025
Source: Release
License
proprietary BytePlus terms
Source: License

API

API model ID
seed-1-6-flash-250715
Source: API model ID

Further details

Notes
Compact multimodal deep-thinking model with a 256K context window; parameter count is not published.
Source: Notes

Prices

Shutdown announced for November 11, 2026. Rates last verified on September 22, 2026.

Choose a price view

1 offer, prices in US dollars per 1M tokens, blended from 3 parts input and 1 part output.

ProviderInputOutputCache readBlended
ByteDancedirect$0.075$0.30$0.015$0.13
ByteDanceabove 128,000 context tokensdirect$0.10$0.80n/a$0.28

Cache prices come from the direct tariff; OpenRouter does not report them for the other providers. Direct price according to the developer, checked Sep 22, 2026. Open the LLM price index

Unchanged since tracking began on Aug 24, 2026.

InputOutput

$0$0.10$0.20$0.30$0.40Aug 24, 2026Oct 19, 2026
Daily observed price in the LLM price index in US dollars per 1M tokens. The index uses a market price from several sources, which can differ from the developer's direct price.

Not included in any subscription in Germany.

All plans in the AI subscription comparison

Frequently asked questions about Seed 1.6 Flash

How much does Seed 1.6 Flash cost?

Through the API $0.075 per 1M input tokens and $0.30 output. That is cheap.

Prices from Oct 19, 2026

Output price since tracking beganDollars per 1M tokens · lower is better

Aug 24, 2026Oct 19, 2026

Unchanged since tracking began on Aug 24, 2026.

How large is the context window of Seed 1.6 Flash?

256,000 tokens. That is how much text Seed 1.6 Flash reads at once, about 380 A4 pages.

Page count as a rule of thumb from Artificial Analysis

Context windowTokens the model reads at once · higher is better
Kimi K2.6262K
Qwen 3.8 27B262K
Kimi K2.7 Code262K
Kimi K2.5262K
Kimi K2 Thinking262K
Mistral Medium 3.5262K
Seed 1.6 FlashRank 76 of 102256K

Can I run Seed 1.6 Flash locally?

No. ByteDance does not publish the weights.

Stated by ByteDance

Open weights
no
License
proprietary BytePlus terms
Parameters
n/a
active per token
n/a

Where can I use Seed 1.6 Flash?

Through the API. Subscription access is not documented in the available plan data.

Through the API

$0.075 input, $0.30 output per 1M tokens, unchanged since Aug 24, 2026.

Sources and data date

Show all 14 sources

Benchmark sources at a glance

Which independent evaluators tested this model. This is coverage, not a quality ranking.

No active result from the reviewed external evaluators can currently be attributed to this model. Other source evidence remains available below.

Without vendor claims and historical or disputed results; those stay labelled in the single values.

TypeEvidence and data date
Price

Base API prices without caching or batch discounts. Higher context tiers and other rates are listed under Costs.

Source: Pricing · API pricing (retrieved October 26, 2026) · BytePlus model access and pricing · Editorial description reviewed October 2, 2026
Input

Documented modalities. Retrieved September 22, 2026.

Source: Model documentation · Model metadata (retrieved September 22, 2026) · BytePlus Seed 1.6 Flash documentation · Editorial description reviewed October 2, 2026
Output

Documented modalities. Retrieved September 22, 2026.

Source: Model documentation · Model metadata (retrieved September 22, 2026) · BytePlus Seed 1.6 Flash documentation · Editorial description reviewed October 2, 2026
Research statusResearch date September 8, 2026. 9 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication.
Not found in the inspected sources

parameters-billions

Original wording

No exact value located in the checked primary model card/API documentation.

Not found in the inspected sources

active-parameters-billions

Original wording

No exact value located in the checked primary model card/API documentation.

Not found in the inspected sources

knowledge-cutoff

Original wording

No exact value located in the checked primary model card/API documentation.

Unresolved

AA benchmark rows

Original wording

No exact Artificial Analysis model page located in the checked public sitemap.

Unresolved

other-table-metric-direction

Note

The same table includes PCS and five phenotype dimensions. This bundle records only aggregate accuracy because the cited report does not establish that the phenotype dimensions are higher-is-better quality scores.

Unresolved

serving-snapshot

Note

The paper names Doubao-Seed-1.6-flash, says the Doubao models were accessed through Volcano Engine API, and links the matching official console route. It does not report the evaluation request date or immutable serving revision. This alias-level result is attached to profile seed-1-6-flash but is not transferred to the separately measured doubao-seed-1-6-flash-250715 checkpoint.

Not published

evaluation-denominator

Note

The paper gives 8,210 as the total curated AEGIS dataset size but does not publish the per-task valid/evaluated sample count for the Seed 1.6 Flash rows. No per-row sample count or confidence interval is assigned.

Additional sourceA Memetic Landscape for Scrutinizing LLMs, OpenReview Table 3 (retrieved October 15, 2026)
Additional sourceAEGIS, ACL 2026, Table 2 (retrieved October 16, 2026)
Additional sourceThe Aggregate, ByteDance model ratings (retrieved October 19, 2026)
Additional sourcen8n AI Benchmark (retrieved October 20, 2026) (unavailable on November 2, 2026) https://n8n.io/ai-benchmark/seed-1-6-flash/
Additional sourceModel metadata (retrieved September 8, 2026)
Additional sourceByteDance Seed 1.6 technical overview · Editorial description reviewed October 2, 2026
Additional sourceOpenRouter: Popularity index (derived), CC BY 4.0
Additional sourceVercel AI Gateway: Popularity index (derived), CC BY 4.0
Additional sourceOllama: cumulative downloads for the popularity index
Additional sourceReplicate: cumulative runs for the popularity index
Additional sourceGoogle Trending Now: limited feed selection for Germany and the US
Additional sourceCreative Commons Attribution 4.0