Seed 2.1 Turbo
ByteDance · Proprietary · bench-align-v5
Capability shape
Seven axes from the ranking source. A missing axis is drawn as a gap.
Runtime service evidence
Measured values with the date they were observed. Nothing is inferred from a sibling model or a provider claim.
Time to first token: latency to the first answer chunk (Artificial Analysis, via BenchLM). Reasoning models include thinking time, so values can run to tens or hundreds of seconds.
Evidence key: ObservedUnavailable
| Measurement | Value | Observed | Last good | Evidence |
|---|---|---|---|---|
| Time to first token | Unavailable | Unavailable | Unavailable | |
| Throughput | 49 tok/s | 2026-09-17 | 2026-09-17 |
Regional or per-endpoint measurements appear only when the API supplies them; none are modelled here.
Endpoint and price matrix
Every published price component, including cache reads and writes.
| Component | USD | Evidence |
|---|---|---|
| Input / 1M tokens | $0.50 | |
| Output / 1M tokens | $2.50 | |
| Cache read / 1M tokens | Unavailable | |
| Cache write / 1M tokens | Unavailable | |
| Blended / 1M (75% input / 25% output) | $1.00 |
Workload-aware monthly cost example
10 conversations per day × 8 messages × 22 active days, 1200 input and 400 output tokens per message, no cache. This uses the same calculator as the cost simulator, so an unavailable applicable rate makes the example unavailable too.
Benchmark record
11 matched benchmark rows with their published value, unit, and provenance.
Knowledge
| Benchmark | Value | Tasks | Difficulty | Provenance |
|---|---|---|---|---|
| SuperGPQA SuperGPQA: Scaling LLM Evaluation Across 285 Graduate Disciplines | 67.4 | 285 disciplines | Graduate level | SuperGPQA: Scaling LLM Evaluation Across 285 Graduate Disciplines |
Coding
| Benchmark | Value | Tasks | Difficulty | Provenance |
|---|---|---|---|---|
| Terminal-Bench 2.1 Terminal-Bench 2.1 (provider run) | 67.6 | Terminal-based software-agent tasks | Professional software engineering | DeepSeek V4 Flash 0731 update |
| NL2Repo NL2Repo | 43.7 | Natural language to repository tasks | System-level software comprehension | MiniMax M2.7: Early Echoes of Self-Evolution |
Agentic
| Benchmark | Value | Tasks | Difficulty | Provenance |
|---|---|---|---|---|
| Terminal-Bench 2.1 Terminal-Bench 2.1 (provider run) | 67.6 | Terminal-based software-agent tasks | Professional software engineering | DeepSeek V4 Flash 0731 update |
Multimodal & Grounded
| Benchmark | Value | Tasks | Difficulty | Provenance |
|---|---|---|---|---|
| MMMU-Pro Massive Multi-discipline Multimodal Understanding Pro | 80.1 | Multimodal academic reasoning | Frontier multimodal | MMMU-Pro: A More Robust Multi-discipline Multimodal Understanding Benchmark |
| Video-MME Video-MME | 89.0 | Video understanding | Broad multimodal video reasoning | Video-MME benchmark |
| MathVision MathVision | 90.1 | Visually grounded math problems | Advanced multimodal mathematics | Qwen3.6 launch benchmarks |
| ERQA ERQA | 71.3 | Evidence-based visual QA | Grounded multimodal reasoning | Qwen3.6 launch benchmarks |
| ZeroBench ZeroBench | 11.0 | 100 visual reasoning questions | Tool-augmented visual reasoning | Muse Spark Eval Methodology |
| CharXiv CharXiv Reasoning | 82.5 | Scientific chart reasoning | Scientific visualization reasoning | CharXiv: Charting Gaps in Realistic Chart Understanding in Multimodal LLMs |
| BabyVision BabyVision | 62.9 | Visual perception tasks | Fine-grained visual perception | Muse Spark 1.1 Evaluation Report |
Lifecycle and limitations log
Lifecycle events the source associates with this model, plus what this profile does not claim.
That is not evidence the model has no lifecycle plan — only that this source published none.