Artificial Analysis Output Speed

1. 1
   Gemini 3.8 Flash

   as “Gemini 3.8 Flash (high)”
   291 tok/s
2. 2
   DeepSeek V4.1 Flash

   as “DeepSeek V4.1 Flash (Reasoning, Max Effort)”
   232 tok/s
3. 3
   Muse Spark 1.3

   as “Muse Spark 1.3 (xhigh)”
   228 tok/s
4. 4
   DeepSeek V4 Flash Vision

   as “DeepSeek V4 Flash Vision (Reasoning, Max Effort)”
   218 tok/s
5. 5
   GPT-5.3 Codex

   as “GPT-5.3 Codex (xhigh)”
   145 tok/s
6. 6
   GPT-6 Luna

   as “GPT-6 Luna (high)”
   134 tok/s
7. 7
   GPT-6 Sol

   as “GPT-6 Sol (medium)”
   114 tok/s
8. 8
   Claude Opus 5.5

   as “Claude Opus 5.5 (Adaptive Reasoning, Xhigh Effort, Default Fallback)”
   92 tok/s
9. 9
   Step 5 Preview
   85 tok/s
10. 10
    GPT-5.6 Terra

    as “GPT-5.6 Terra (max)”
    83 tok/s
11. 11
    Claude Sonnet 5

    as “Claude Sonnet 5 (Adaptive Reasoning, Max Effort)”
    79 tok/s
12. 12
    DeepSeek V4 Pro 0813

    as “DeepSeek V4 Pro 0813 (Reasoning, Max Effort)”
    66 tok/s
13. 13
    Claude Fable 5.1

    as “Claude Fable 5.1 (Adaptive Reasoning, Max Effort, Default Fallback)”
    66 tok/s
14. 14
    Grok 4.6

    as “Grok 4.6 (high)”
    61 tok/s
15. 15
    GLM-5.3

    as “GLM-5.3 (max)”
    57 tok/s
16. 16
    Qwen3.8-Flash-Next
    54 tok/s
17. 17
    GPT-6 Astra

    as “GPT-6 Astra (max)”
    52 tok/s
18. 18
    Grok 4.7

    as “Grok 4.7 (high)”
    52 tok/s
19. 19
    MiMo-V2.6-Pro
    49 tok/s
20. 20
    GLM 5.3 Flash
    46 tok/s
21. 21
    Qwen3.8 27B

    as “Qwen3.8 27B (xhigh)”
    43 tok/s
22. 22
    Qwen3.8 Max (0902)
    39 tok/s
23. 23
    Kimi K3

    as “Kimi K3 (low)”
    38 tok/s
24. 24
    Qwen3.8 2.4T A95B
    38 tok/s

History builds with every daily capture. One capture so far.