Skip to content
MCS · Models · 62

Model Ratings.

The tracked cohort, scored and tiered.

62 models ranked against the coding muscle recipe · every axis sourced.

Scores are relative to the tracked cohort — swap the recipe or the field and the ranks move. The weights are an editorial call; the axis values underneath are sourced, cell by cell.

Snapshot · verified 1w ago · 2026-10-05

Recipe

Ship-real-PRs ability — agentic + competitive coding weighted hardest.

The board · ranked · Coding muscle

Each cell brightens toward the cohort leader and is ringed when it holds it; dimmed rows are sparse (low coverage) — read their composite with care.

  1. 1
    Claude Opus 5.5
    Anthropic
    25% · lowS100
    1. SWE-bench—
    2. LiveCode—
    3. AA Index58
    4. GPQA—

    Strongest: aa_index · Weakest: aa_index

  2. 2
    Claude Sonnet 5.5
    Anthropic
    25% · lowS94
    1. SWE-bench—
    2. LiveCode—
    3. AA Index56
    4. GPQA—

    Strongest: aa_index · Weakest: aa_index

  3. 3
    Claude Opus 5
    Anthropic
    75% · highS93
    1. SWE-bench96%
    2. LiveCode—
    3. AA Index51
    4. GPQA93.2%

    Strongest: swebench · Weakest: aa_index

  4. 4
    GPT-6 Astra
    OpenAI
    50% · mediumS92
    1. SWE-bench—
    2. LiveCode—
    3. AA Index53
    4. GPQA96.1%

    Strongest: aa_index · Weakest: aa_index

  5. 5
    Claude Fable 5.1
    Anthropic
    50% · mediumS90
    1. SWE-bench—
    2. LiveCode—
    3. AA Index53
    4. GPQA93.7%

    Strongest: aa_index · Weakest: aa_index

  6. 6
    Claude Fable 5
    Anthropic
    75% · highS90
    1. SWE-bench95%
    2. LiveCode—
    3. AA Index50
    4. GPQA92.6%

    Strongest: swebench · Weakest: aa_index

  7. 7
    Gemini 4 Argon
    Google DeepMind
    25% · lowS86
    1. SWE-bench—
    2. LiveCode—
    3. AA Index53
    4. GPQA—

    Strongest: aa_index · Weakest: aa_index

  8. 8
    GPT-6.1 Sol
    OpenAI
    25% · lowA83
    1. SWE-bench—
    2. LiveCode—
    3. AA Index52
    4. GPQA—

    Strongest: aa_index · Weakest: aa_index

  9. 9
    Muse Spark 1.3
    Meta
    50% · mediumA82
    1. SWE-bench—
    2. LiveCode—
    3. AA Index48
    4. GPQA93.5%

    Strongest: gpqa · Weakest: aa_index

  10. 10
    GPT-5.6 Sol
    OpenAI
    50% · mediumA81
    1. SWE-bench—
    2. LiveCode—
    3. AA Index47
    4. GPQA94.1%

    Strongest: gpqa · Weakest: aa_index

  11. 11
    Qwen3.8-Flash-Next
    Alibaba (Qwen)
    75% · highA78
    1. SWE-bench—
    2. LiveCode91.9%
    3. AA Index40
    4. GPQA91.7%

    Strongest: livecodebench · Weakest: aa_index

  12. 12
    Grok 4.6
    xAI
    50% · mediumA77
    1. SWE-bench—
    2. LiveCode—
    3. AA Index44
    4. GPQA94.9%

    Strongest: gpqa · Weakest: aa_index

  13. 13
    Qwen3.8 Max (0902)
    Alibaba (Qwen)
    50% · mediumA77
    1. SWE-bench—
    2. LiveCode—
    3. AA Index45
    4. GPQA92.8%

    Strongest: gpqa · Weakest: aa_index

  14. 14
    GLM-5.3
    Z.AI (Zhipu)
    50% · mediumA76
    1. SWE-bench—
    2. LiveCode—
    3. AA Index45
    4. GPQA91.7%

    Strongest: gpqa · Weakest: aa_index

  15. 15
    Kimi K3
    Moonshot AI
    50% · mediumA76
    1. SWE-bench—
    2. LiveCode—
    3. AA Index44
    4. GPQA93.5%

    Strongest: gpqa · Weakest: aa_index

  16. 16
    Step 5 Preview
    StepFun
    50% · mediumA76
    1. SWE-bench—
    2. LiveCode—
    3. AA Index44
    4. GPQA93.5%

    Strongest: gpqa · Weakest: aa_index

  17. 17
    Gemini 3.8 Flash
    Google DeepMind
    50% · mediumA73
    1. SWE-bench—
    2. LiveCode—
    3. AA Index41
    4. GPQA95.3%

    Strongest: gpqa · Weakest: aa_index

  18. 18
    GPT-5.6 Terra
    OpenAI
    50% · mediumA72
    1. SWE-bench—
    2. LiveCode—
    3. AA Index42
    4. GPQA92.5%

    Strongest: gpqa · Weakest: aa_index

  19. 19
    GPT-6 Sol
    OpenAI
    25% · lowA72
    1. SWE-bench—
    2. LiveCode—
    3. AA Index48
    4. GPQA—

    Strongest: aa_index · Weakest: aa_index

  20. 20
    GLM-5.3-Flash
    Z.AI (Zhipu)
    50% · mediumA71
    1. SWE-bench—
    2. LiveCode—
    3. AA Index42
    4. GPQA91.2%

    Strongest: gpqa · Weakest: aa_index

  21. 21
    Claude Opus 4.8
    Anthropic
    75% · highA71
    1. SWE-bench88.6%
    2. LiveCode—
    3. AA Index42
    4. GPQA93.6%

    Strongest: swebench · Weakest: aa_index

  22. 22
    gpt-oss-120b
    OpenAI
    25% · lowA71
    1. SWE-bench—
    2. LiveCode—
    3. AA Index—
    4. GPQA80.1%

    Strongest: gpqa · Weakest: gpqa

  23. 23
    Gemini 3.7 Flash
    Google DeepMind
    50% · mediumB69
    1. SWE-bench—
    2. LiveCode—
    3. AA Index39
    4. GPQA94.5%

    Strongest: gpqa · Weakest: aa_index

  24. 24
    Qwen3.8 Max
    Alibaba (Qwen)
    50% · mediumB69
    1. SWE-bench—
    2. LiveCode—
    3. AA Index40
    4. GPQA92.6%

    Strongest: gpqa · Weakest: aa_index

  25. 25
    Claude Opus 4.7
    Anthropic
    75% · highB68
    1. SWE-bench87.6%
    2. LiveCode—
    3. AA Index41
    4. GPQA94.2%

    Strongest: swebench · Weakest: aa_index

  26. 26
    GPT-5.5
    OpenAI
    75% · highB68
    1. SWE-bench88.7%
    2. LiveCode—
    3. AA Index38
    4. GPQA93.5%

    Strongest: swebench · Weakest: aa_index

  27. 27
    Qwen3.8-27B
    Alibaba (Qwen)
    75% · highB68
    1. SWE-bench—
    2. LiveCode90.3%
    3. AA Index34
    4. GPQA90.5%

    Strongest: livecodebench · Weakest: aa_index

  28. 28
    Grok 4.7
    xAI
    25% · lowB67
    1. SWE-bench—
    2. LiveCode—
    3. AA Index46
    4. GPQA—

    Strongest: aa_index · Weakest: aa_index

  29. 29
    MiMo-V2.6-Pro
    Xiaomi
    25% · lowB67
    1. SWE-bench—
    2. LiveCode—
    3. AA Index46
    4. GPQA—

    Strongest: aa_index · Weakest: aa_index

  30. 30
    DeepSeek V4.1 Flash
    DeepSeek
    50% · mediumB66
    1. SWE-bench—
    2. LiveCode—
    3. AA Index39
    4. GPQA90.9%

    Strongest: gpqa · Weakest: aa_index

  31. 31
    GPT-5.6 Luna
    OpenAI
    50% · mediumB64
    1. SWE-bench—
    2. LiveCode—
    3. AA Index37
    4. GPQA91.1%

    Strongest: gpqa · Weakest: aa_index

  32. 32
    DeepSeek V4 Pro 0813
    DeepSeek
    50% · mediumB63
    1. SWE-bench—
    2. LiveCode—
    3. AA Index36
    4. GPQA92.8%

    Strongest: gpqa · Weakest: aa_index

  33. 33
    DeepSeek V4 Pro
    DeepSeek
    100% · highB60
    1. SWE-bench80.6%
    2. LiveCode93.5%
    3. AA Index30
    4. GPQA90.1%

    Strongest: livecodebench · Weakest: aa_index

  34. 34
    DeepSeek V4 Flash 0731
    DeepSeek
    50% · mediumB59
    1. SWE-bench—
    2. LiveCode—
    3. AA Index34
    4. GPQA91%

    Strongest: gpqa · Weakest: aa_index

  35. 35
    gpt-oss-20b
    OpenAI
    25% · lowB55
    1. SWE-bench—
    2. LiveCode—
    3. AA Index—
    4. GPQA71.5%

    Strongest: gpqa · Weakest: gpqa

  36. 36
    MiniMax-M3
    MiniMax
    50% · mediumC53
    1. SWE-bench—
    2. LiveCode—
    3. AA Index29
    4. GPQA92.9%

    Strongest: gpqa · Weakest: aa_index

  37. 37
    Kimi K2.6
    Moonshot AI
    100% · highC50
    1. SWE-bench80.2%
    2. LiveCode89.6%
    3. AA Index27
    4. GPQA90.5%

    Strongest: livecodebench · Weakest: aa_index

  38. 38
    DeepSeek V4 Flash (Preview)
    DeepSeek
    100% · highC50
    1. SWE-bench79%
    2. LiveCode91.6%
    3. AA Index24
    4. GPQA88.1%

    Strongest: livecodebench · Weakest: aa_index

  39. 39
    Grok 4.5
    xAI
    25% · lowC47
    1. SWE-bench—
    2. LiveCode—
    3. AA Index39
    4. GPQA—

    Strongest: aa_index · Weakest: aa_index

  40. 40
    Gemini 3.1 Pro
    Google DeepMind
    75% · highC45
    1. SWE-bench80.6%
    2. LiveCode—
    3. AA Index30
    4. GPQA94.1%

    Strongest: gpqa · Weakest: aa_index

  41. 41
    GPT-6 Luna
    OpenAI
    25% · lowC44
    1. SWE-bench—
    2. LiveCode—
    3. AA Index38
    4. GPQA—

    Strongest: aa_index · Weakest: aa_index

  42. 42
    MiMo-V2.6-Flash
    Xiaomi
    25% · lowC44
    1. SWE-bench—
    2. LiveCode—
    3. AA Index38
    4. GPQA—

    Strongest: aa_index · Weakest: aa_index

  43. 43
    GLM-5.1
    Z.AI (Zhipu)
    50% · mediumC43
    1. SWE-bench—
    2. LiveCode—
    3. AA Index26
    4. GPQA86.2%

    Strongest: gpqa · Weakest: aa_index

  44. 44
    DeepSeek V4 Flash Vision Exp
    DeepSeek
    25% · lowD36
    1. SWE-bench—
    2. LiveCode—
    3. AA Index35
    4. GPQA—

    Strongest: aa_index · Weakest: aa_index

  45. 45
    Gemini 3.6 Flash
    Google DeepMind
    25% · lowD33
    1. SWE-bench—
    2. LiveCode—
    3. AA Index34
    4. GPQA—

    Strongest: aa_index · Weakest: aa_index

  46. 46
    GLM-5.2
    Z.AI (Zhipu)
    25% · lowD33
    1. SWE-bench—
    2. LiveCode—
    3. AA Index34
    4. GPQA—

    Strongest: aa_index · Weakest: aa_index

  47. 47
    Gemini 3.5 Flash
    Google DeepMind
    25% · lowD31
    1. SWE-bench—
    2. LiveCode—
    3. AA Index33
    4. GPQA—

    Strongest: aa_index · Weakest: aa_index

  48. 48
    GPT-5.3 Codex
    OpenAI
    25% · lowD31
    1. SWE-bench—
    2. LiveCode—
    3. AA Index33
    4. GPQA—

    Strongest: aa_index · Weakest: aa_index

  49. 49
    Claude Sonnet 4.6
    Anthropic
    50% · mediumD28
    1. SWE-bench80.2%
    2. LiveCode—
    3. AA Index30
    4. GPQA—

    Strongest: swebench · Weakest: aa_index

  50. 50
    Qwen3.7 Max
    Alibaba (Qwen)
    75% · highD28
    1. SWE-bench—
    2. LiveCode78.2%
    3. AA Index29
    4. GPQA92.4%

    Strongest: gpqa · Weakest: livecodebench

  51. 51
    Phi-4
    Microsoft
    25% · lowD27
    1. SWE-bench—
    2. LiveCode—
    3. AA Index—
    4. GPQA56.1%

    Strongest: gpqa · Weakest: gpqa

  52. 52
    Muse Spark
    Meta
    25% · lowD25
    1. SWE-bench—
    2. LiveCode—
    3. AA Index31
    4. GPQA—

    Strongest: aa_index · Weakest: aa_index

  53. 53
    Claude Sonnet 5
    Anthropic
    50% · mediumD16
    1. SWE-bench72.7%
    2. LiveCode—
    3. AA Index38
    4. GPQA—

    Strongest: aa_index · Weakest: swebench

  54. 54
    Kimi K2.7 Code
    Moonshot AI
    25% · lowD11
    1. SWE-bench—
    2. LiveCode—
    3. AA Index26
    4. GPQA—

    Strongest: aa_index · Weakest: aa_index

  55. 55
    MiMo-V2.5-Pro
    Xiaomi
    25% · lowD11
    1. SWE-bench—
    2. LiveCode—
    3. AA Index26
    4. GPQA—

    Strongest: aa_index · Weakest: aa_index

  56. 56
    Mistral Small 3.2 24B
    Mistral AI
    25% · lowD9
    1. SWE-bench—
    2. LiveCode—
    3. AA Index—
    4. GPQA46.1%

    Strongest: gpqa · Weakest: gpqa

  57. 57
    Grok 4.3
    xAI
    25% · lowD8
    1. SWE-bench—
    2. LiveCode—
    3. AA Index25
    4. GPQA—

    Strongest: aa_index · Weakest: aa_index

  58. 58
    Qwen3.7 Plus
    Alibaba (Qwen)
    25% · lowD8
    1. SWE-bench—
    2. LiveCode—
    3. AA Index25
    4. GPQA—

    Strongest: aa_index · Weakest: aa_index

  59. 59
    GPT-5.4 mini
    OpenAI
    25% · lowD6
    1. SWE-bench—
    2. LiveCode—
    3. AA Index24
    4. GPQA—

    Strongest: aa_index · Weakest: aa_index

  60. 60
    Gemini 3.5 Flash-Lite
    Google DeepMind
    25% · lowD0
    1. SWE-bench—
    2. LiveCode—
    3. AA Index22
    4. GPQA—

    Strongest: aa_index · Weakest: aa_index

  61. 61
    Gemma 3 27B
    Google
    25% · lowD0
    1. SWE-bench—
    2. LiveCode—
    3. AA Index—
    4. GPQA41.4%

    Strongest: gpqa · Weakest: gpqa

  62. 62
    Qwen3-32B
    Alibaba (Qwen)
    0% · lowD0
    1. SWE-bench—
    2. LiveCode—
    3. AA Index—
    4. GPQA—

    Strongest: — · Weakest: —