gemini/gemini-3.5-flash| Provider | Input, $ per 1M | Output, $ per 1M | In our data since |
|---|---|---|---|
| unorouter | $0.186 | $1.114 | 31 Jul 2026 |
| neon | $1.5 | $9 | 31 Jul 2026 |
| AIHubMix | $1.5 | $9 | 31 Jul 2026 |
| Abacus.AI | $1.5 | $9 | 31 Jul 2026 |
| CrossModel | $1.5 | $9 | 31 Jul 2026 |
| FastRouter | $1.5 | $9 | 31 Jul 2026 |
| GitHub Copilot | $1.5 | $9 | 31 Jul 2026 |
| Google DeepMind | $1.5 | $9 | 31 Jul 2026 |
| Google Vertex AI | $1.5 | $9 | 31 Jul 2026 |
| Kilo Code | $1.5 | $9 | 31 Jul 2026 |
| LLM Gateway | $1.5 | $9 | 31 Jul 2026 |
| Merge Gateway | $1.5 | $9 | 31 Jul 2026 |
| NEAR AI | $1.5 | $9 | 31 Jul 2026 |
| NanoGPT | $1.5 | $9 | 31 Jul 2026 |
| OpenCode Zen | $1.5 | $9 | 31 Jul 2026 |
| OpenRouter | $1.5 | $9 | 31 Jul 2026 |
| Pioneer | $1.5 | $9 | 31 Jul 2026 |
| SAP AI Core | $1.5 | $9 | 31 Jul 2026 |
| Vercel | $1.5 | $9 | 31 Jul 2026 |
| ZenMux | $1.5 | $9 | 31 Jul 2026 |
| Poe | $1.515 | $9.091 | 31 Jul 2026 |
| xpersona | $1.55 | $12.2 | 31 Jul 2026 |
| Venice AI | $1.55 | $9.45 | 31 Jul 2026 |
Only the base tier and only per-token prices are shown. Batch, discounted and cached rates, as well as prices per image or per second of video, do not go into this table: they cannot stand in the same column as a price per million tokens.
The date in the last column is the day this price first entered our collection. The price may well be older: before that day we simply were not recording it. It has not changed since — otherwise a new row with a new date would stand in its place.
| Task set | Result | Run conditions | Measured by |
|---|---|---|---|
| Arena Score, mathematics | 1,523.8 1,498–1,549 | effort: high | — |
| Arena Score, expert questions | 1,509.47 1,491–1,528 | effort: high | — |
| Arena Score in Russian | 1,499.47 1,480–1,519 | effort: high | — |
| Arena Score, programming | 1,492.01 1,480–1,504 | effort: high | — |
| Arena Score, multi-turn dialogue | 1,489.28 1,475–1,504 | effort: high | — |
| Arena Score, web development | 1,486.18 1,478–1,494 | effort: medium | — |
| Arena Score, hard prompts | 1,485.27 1,477–1,493 | effort: high | — |
| Arena Score, long queries | 1,481.88 1,472–1,492 | effort: high | — |
| Arena Score, overall | 1,480.35 1,474–1,487 | effort: high | — |
| Arena Score in English | 1,479.43 1,470–1,488 | effort: high | — |
| Arena Score in Spanish | 1,471.19 1,440–1,502 | effort: high | — |
| Arena Score, creative writing | 1,470.09 1,454–1,486 | effort: high | — |
| Arena Score, instruction following | 1,464.95 1,454–1,476 | effort: high | — |
| Arena Score in French | 1,460.6 1,427–1,494 | effort: high | — |
| Arena Score, understanding diagrams | 1,304.8 1,283–1,326 | effort: high | — |
| Arena Score, text recognition in images | 1,303.5 1,290–1,317 | effort: high | — |
| Arena Score, working with images | 1,301.48 1,289–1,314 | effort: high | — |
| Mock AIME 2024–2025 — olympiad problems | 95.55 % | effort: minimal · with a tuned harness | Epoch evaluations |
| ARC-AGI — generalising to unseen patterns | 92.5 % | effort: high · with a tuned harness | https://arcprize.org/leaderboard |
| GPQA Diamond — graduate-level questions | 90.4 % | effort: minimal · with a tuned harness | Epoch evaluations |
| SWE-bench Verified — fixing bugs in repositories | 79.34 % | effort: high · with a tuned harness | Epoch evaluations |
| ARC-AGI-2 | 72.08 % | effort: high | — |
| SimpleBench — trick questions | 72.04 % | — | SimpleBench Leaderboard |
| SimpleQA Verified — factual accuracy | 68.4 % | effort: high | Epoch evaluations |
| FrontierMath, levels 1–3 | 62.81 % | effort: high | Epoch evaluations |
| WeirdML — unusual machine learning tasks | 62.64 % | effort: high | https://htihle.github.io/weirdml.html |
| CursorBench — edits in the editor | 49.8 % | — | https://cursor.com/cursorbench |
| APEX-Agents | 49.6 % | — | — self-reported |
| Chess puzzles | 47.39 % | effort: minimal | Epoch evaluations |
| FrontierMath, level 4 — research-grade problems | 26.83 % | effort: high | Epoch evaluations |
| CritPt — physics problems | 13.14 % | effort: high | — |
| Arena Score, agent tasks | -0.01 0–0 | effort: high | — |