dots3-note Preview
Dots StudioOpen weightsReasoningProvisionalReleased Aug 14, 2026Input: text, image
Overall score
64.1
Rank 41 of 222 · 49.7–78.4
Context window
512K
512,000 tokens
Price per 1M tokens
— in · — out
No price listed
Blended price
—
3:1 input:output tokens
Output speed
—
Not measured yet
Category scores
Scores per capability area with the model's rank among ranked models in that category.
- 1Instruction following#3289.1
- 2Multimodal79.6
- 3Knowledge78.0
- 4Reasoning76.6
- 5Agentic & tool use71.1
- 6Coding57.2
Benchmark results (29)
As reported in system cards, papers and public evaluations. Position compares this model with every other model reporting the same benchmark.
Multimodal
| MMMU-Pro | 79.1 | #14 of 48 |
| MathVision | 87.7 | #10 of 20 |
| SimpleVQA | 72.5 | #6 of 17 |
| CharXiv w/o tools | 83.1 | #9 of 14 |
| VideoMMMU | 86.8 | #3 of 12 |
| ZeroBench | 19.0 | #7 of 8 |
| BabyVision | 50.0 | #7 of 8 |
| MMVU | 79.9 | #4 of 7 |
| PerceptionBench | 53.4 | #3 of 3 |
| GDP.pdf (no tools) | 60.7 | #2 of 2 |
Agentic & tool use
| Terminal-Bench 2.1 | 75.1 | #35 of 68 |
| BrowseComp | 83.3 | #19 of 50 |
| Claw-Eval | 73.4 | #5 of 36 |
| HLE w/ tools | 52.6 | #18 of 27 |
| Toolathlon-Verified | 55.6 | #18 of 22 |
| DeepSearchQA | 92.1 | #8 of 20 |
| WideResearch | 78.9 | #7 of 16 |
| APEX-Agents | 30.8 | #6 of 8 |
Coding
| SWE-bench Verified | 78.4 | #24 of 98 |
| SWE-bench Pro | 61.0 | #25 of 69 |
| LiveCodeBench v6 | 91.5 | #5 of 46 |
| NL2Repo | 49.8 | #10 of 31 |
| Codeforces | 3056.0 | #3 of 4 |
Instruction following
| IFEval | 93.9 | #7 of 53 |
| IFBench | 80.4 | #8 of 42 |
Multilingual
| SWE Multilingual | 75.7 | #25 of 46 |
Reasoning
| ARC-AGI-2 | 81.4 | #16 of 41 |
Knowledge
| HLE | 52.6 | #14 of 69 |
Math
| IMOAnswerBench | 90.9 | #1 of 9 |
Pricing by source
USD per 1M tokens.
| Source | Input | Output | Cached in |
|---|---|---|---|
| List price | — | — | — |
| OpenRouter dots-studio/dots-3-note-preview:free | — | — | — |
Dots Studio publishes the BF16 and FP8 dots3-note Preview checkpoints under Apache-2.0 for self-hosting and does not publish a first-party hosted token rate for the exact model. BenchLM represents the open-weight row as self-host/free-per-token before infrastructure costs.
Speed
Measured by us through OpenRouter (median of trials).
Not measured yet.
Related models
Similar score
- GPT-5.6 Luna64.4 · $0.45
- MiMo-V2.6-Flash64.6 · $0.17
- Gemini 3.1 Pro64.7 · $4.50
- Grok 4.563.3 · $3.00
- Claude Sonnet 564.9 · $4.00