Speed
Output speed
26 models with a published score, best first. Higher is better. The record has been broken 3 times since Oct 20, 2026.
239Claude Haiku 5.5Rank 1, published Oct 20, 2026
Leaderboard
Each model's latest published score. Every score links to its source.
| # | Model | Tok / s | Date | Note | Source |
|---|---|---|---|---|---|
| 1 | A Claude Haiku 5.5RecordAnthropic | 239 | Oct 20, 2026 | Max | artificialanalysis.ai |
| 2 | D DeepSeek-V4.1-FlashDeepSeek | 235 | Oct 20, 2026 | Max | artificialanalysis.ai |
| 3 | I Ling 3.1 FlashInclusionAI | 218 | Oct 20, 2026 | artificialanalysis.ai | |
| 4 | D DeepSeek V4.1 Flash 1017DeepSeek | 212 | Oct 20, 2026 | Max | artificialanalysis.ai |
| 5 | M Muse WonderMeta | 164 | Oct 20, 2026 | Max | artificialanalysis.ai |
| 6 | O GPT-5.6 TerraOpenAI | 159 | Oct 20, 2026 | Max | artificialanalysis.ai |
| 7 | G Gemini 3.8 FlashGoogle | 146 | Oct 20, 2026 | High | artificialanalysis.ai |
| 8 | A Claude Sonnet 5.5Anthropic | 140 | Oct 20, 2026 | Max, Default Fallback | artificialanalysis.ai |
| 8 | M Muse Spark OpenMeta | 140 | Oct 20, 2026 | Xhigh | artificialanalysis.ai |
| 10 | M Muse Spark 1.3Meta | 123 | Oct 20, 2026 | Max | artificialanalysis.ai |
| 11 | A Claude Opus 5.5Anthropic | 96 | Oct 20, 2026 | Max, Default Fallback | artificialanalysis.ai |
| 11 | M MiniMax M3.1 FlashMiniMax | 96 | Oct 20, 2026 | artificialanalysis.ai | |
| 13 | S Step 5 PreviewStepFun | 86 | Oct 20, 2026 | artificialanalysis.ai | |
| 14 | G Gemini 4 ArgonGoogle | 82 | Oct 20, 2026 | High | artificialanalysis.ai |
| 15 | Z GLM-5.3Zhipu | 78 | Oct 20, 2026 | Max | artificialanalysis.ai |
| 15 | Z GLM-5.3-FlashZhipu | 78 | Oct 20, 2026 | artificialanalysis.ai | |
| 17 | X Grok 4.7xAI | 77 | Oct 20, 2026 | Xhigh | artificialanalysis.ai |
| 18 | A Claude Fable 5.1Anthropic | 72 | Oct 20, 2026 | Max, Default Fallback | artificialanalysis.ai |
| 18 | O GPT-6 AstraOpenAI | 72 | Oct 20, 2026 | Max | artificialanalysis.ai |
| 20 | M Mistral Large 4 PreviewMistral | 56 | Oct 20, 2026 | artificialanalysis.ai | |
| 21 | O GPT-6.1 SolOpenAI | 55 | Oct 20, 2026 | Max | artificialanalysis.ai |
| 21 | A Qwen3.8-Flash-NextAlibaba | 55 | Oct 20, 2026 | artificialanalysis.ai | |
| 23 | X MiMo-V2.6-ProXiaomi | 41 | Oct 20, 2026 | artificialanalysis.ai | |
| 23 | M Kimi K3Moonshot | 41 | Oct 20, 2026 | Max | artificialanalysis.ai |
| 25 | A Qwen3.8-MaxAlibaba | 38 | Oct 20, 2026 | 0902 | artificialanalysis.ai |
| 25 | A Qwen3.8 2.4T A95BAlibaba | 38 | Oct 20, 2026 | artificialanalysis.ai |
Record history
The best score ever published on Output speed, and who held it. Each step is a new record.
| Date | Record holder | Tok / s | Gain | Note |
|---|---|---|---|---|
| Oct 20, 2026 | Claude Haiku 5.5Current | 239 | +27 | Max |
| Oct 20, 2026 | DeepSeek V4.1 Flash 1017 | 212 | +48 | Max |
| Oct 20, 2026 | Muse Wonder | 164 | +3 | Max |
| Oct 20, 2026 | Muse Wonder | 161 | First |
Method
Median output tokens per second, Artificial Analysis. Higher is better. Board source: artificialanalysis.ai.
We collect the scores the board and the labs publish and run none ourselves. Ranks use each model's latest reading. 32 readings stored for this board, the newest from Oct 20, 2026.