Gemini 3.5 Flash
- Input price
- —
- Per 1 million tokens (USD)
- Output price
- —
- Per 1 million tokens (USD)
- Effort levels
- not collected
- Benchmark results
- 7total
- Benchmark types: 7
Model information
- API model ID
- gemini-3.5-flash
- release date
- 2026-05-19
- effort
- not collected
- Effort source
- Not collected: the Gemini API model page (aistudio.google.com/docs/models/gemini-3.5-flash) only says Thinking is Supported, with no list of levels, and it is absent from the table in the thinking documentation (discontinued; API automatically redirects to gemini-3.6-flash).
- Claude Code
- unavailable
- Codex
- unavailable
- Availability notes
- Both are false because the model appears in neither the Claude Code official model list (code.claude.com/docs/en/model-config alias table, retrieved 2026-11-13; original: the table for "Use a model alias to select model settings without remembering exact version numbers" contains only Anthropic models such as fable, sonnet, opus, and haiku) nor the Codex official model list (learn.chatgpt.com/docs/models, retrieved 2026-11-13; original: "Recommended models" contains only the GPT-6 series) (connections via a custom provider are not counted as official options, and connections via a gateway are not counted as official options).
- Availability source
- https://code.claude.com/docs/en/model-config
Price (USD per 1 million tokens)
Prices do not vary by effort; token consumption does.
| input | — |
|---|---|
| output | — |
| cache read | — |
| cache write | — |
Discontinued (API redirected to gemini-3.6-flash). Not collected.
Benchmark results
| Benchmark | effort | Score | Source cost (metric) | measured/reported by | source | Conditions and extraction method |
|---|---|---|---|---|---|---|
| Terminal-Bench 2.1not stated | not stated | 76.2% | — | Gemini 3.5 Flash model cardofficial · Retrieved 2026-11-13 | Vendor-reported value. Table in the Google model card. methodology PDF (deepmind.google/models/evals-methodology/gemini-3-5-flash): All of the results are all run with the Gemini API ... with default sampling settings unless indicated otherwise. 3.5 Flash and 3 Flash computed in-house with the Terminus 2 harness (3.1 Pro is not used because its value is from the leaderboard). thinking level not statedtranscribed from tableSource excerptCoding | Terminal-bench 2.1 Agentic terminal coding | Terminus-2 harness | 76.2%Location Gemini 3.5 Flash model card Evaluation table, Terminal-bench 2.1 row, Gemini 3.5 Flash column. In-house computation basis: methodology PDF, Coding / Terminal-Bench 2.1 | |
| CharXiv Reasoningnot stated | not stated | 84.2% | — | Gemini 3.5 Flash model cardofficial · Retrieved 2026-11-13 | Vendor-reported value. Table in the Google model card. methodology PDF (deepmind.google/models/evals-methodology/gemini-3-5-flash): All of the results are all run with the Gemini API ... with default sampling settings unless indicated otherwise. Gemini computed in-house; without tools. thinking level not statedtranscribed from tableSource excerptMultimodal | CharXiv Reasoning Information synthesis from complex charts | No tools | 84.2%Location Gemini 3.5 Flash model card Evaluation table, CharXiv Reasoning(No tools) row, Gemini 3.5 Flash column. In-house computation basis: methodology PDF, Image / CharXiv Reasoning | |
| SWE-Bench Pronot stated | not stated | 55.1% | — | Gemini 3.5 Flash model cardofficial · Retrieved 2026-11-13 | Vendor-reported value. Table in the Google model card. methodology PDF (deepmind.google/models/evals-methodology/gemini-3-5-flash): All of the results are all run with the Gemini API ... with default sampling settings unless indicated otherwise. Gemini computed in-house; average of 5 runs; internal version of the Antigravity harness; single trial. thinking level not statedtranscribed from tableSource excerptSWE-Bench Pro (Public) Diverse agentic coding tasks | Single attempt | 55.1%Location Gemini 3.5 Flash model card Evaluation table, SWE-Bench Pro (Public) row, Gemini 3.5 Flash column. In-house computation basis: methodology PDF, Coding / SWE-Bench Pro | |
| OSWorld-Verifiednot stated | not stated | 78.4% | — | Gemini 3.5 Flash model cardofficial · Retrieved 2026-11-13 | Vendor-reported value. Table in the Google model card. methodology PDF (deepmind.google/models/evals-methodology/gemini-3-5-flash): All of the results are all run with the Gemini API ... with default sampling settings unless indicated otherwise. Gemini computed in-house; average of 5 runs (1 trial each); up to 100 steps. thinking level not statedtranscribed from tableSource excerptUI Control | OSWorld-Verified Agentic computer use | 78.4%Location Gemini 3.5 Flash model card Evaluation table, OSWorld-Verified row, Gemini 3.5 Flash column. In-house computation basis: methodology PDF, UI Control / OSWorld-Verified | |
| MMMU-Pronot stated | not stated | 83.6% | — | Gemini 3.5 Flash model cardofficial · Retrieved 2026-11-13 | Vendor-reported value. Table in the Google model card. methodology PDF (deepmind.google/models/evals-methodology/gemini-3-5-flash): All of the results are all run with the Gemini API ... with default sampling settings unless indicated otherwise. Gemini computed in-house; average of Standard (10 choices) and Vision; without tools. thinking level not statedtranscribed from tableSource excerptMMMU-Pro Multimodal understanding and reasoning | No tools | 83.6%Location Gemini 3.5 Flash model card Evaluation table, MMMU-Pro(No tools) row, Gemini 3.5 Flash column. In-house computation basis: methodology PDF, Image / MMMU-Pro | |
| MRCR v2 (8-needle, 128k)not stated | not stated | 77.3% | — | Gemini 3.5 Flash model cardofficial · Retrieved 2026-11-13 | Vendor-reported value. Table in the Google model card. methodology PDF (deepmind.google/models/evals-methodology/gemini-3-5-flash): All of the results are all run with the Gemini API ... with default sampling settings unless indicated otherwise. All models computed in-house; cumulative score up to 128k. thinking level not statedtranscribed from tableSource excerptLong context | MRCR v2 (8-needle) Long context performance | 128k (average) | 77.3%Location Gemini 3.5 Flash model card Evaluation table, MRCR v2 (8-needle) 128k (average) row, Gemini 3.5 Flash column. In-house computation basis: methodology PDF, Long Context / MRCR v2 | |
| MRCR v2 (8-needle, 1M)not stated | not stated | 26.6% | — | Gemini 3.5 Flash model cardofficial · Retrieved 2026-11-13 | Vendor-reported value. Table in the Google model card. methodology PDF (deepmind.google/models/evals-methodology/gemini-3-5-flash): All of the results are all run with the Gemini API ... with default sampling settings unless indicated otherwise. All models computed in-house; pointwise value at 1M. thinking level not statedtranscribed from tableSource excerpt1M (pointwise) | 26.6%Location Gemini 3.5 Flash model card Evaluation table, MRCR v2 (8-needle) 1M (pointwise) row, Gemini 3.5 Flash column. In-house computation basis: methodology PDF, Long Context / MRCR v2 |
Related rankings
- Overall ranking (all models, records)
- Terminal-Bench 2.1 ranking by metric (all models, records)
- CharXiv Reasoning ranking by metric (all models, records)
- SWE-Bench Pro ranking by metric (all models, records)
- OSWorld-Verified ranking by metric (all models, records)
- MMMU-Pro ranking by metric (all models, records)
- MRCR v2 (8-needle, 128k) ranking by metric (all models, records)
- MRCR v2 (8-needle, 1M) ranking by metric (all models, records)
- Recommendations by role