Skip to content

Scenario guide

Best AI models for Local / Self-Hosted Coding

As of October 25, 2026, the highest-ranked model for Local / Self-Hosted Coding is Qwen2.5 72B Instruct, with a scenario score of 92.8 out of 100. The score blends this scenario's benchmark weights with its cost weight.

A coding assistant you run on your own hardware — no API bills, full privacy, works offline. Restricted to open-weight, self-hostable models and weighted toward practical coding and agentic ability, with a slice for the self-correction that smaller local models lean on inside a good harness. There is no cost axis because inference runs on hardware you already own.

Rankings use the same scenario weights and cost blending as the interactive leaderboard on AI Model Analyzer. Data is min-max normalised per benchmark; missing scores are skipped without penalty.

  1. 1
    Qwen2.5 72B Instruct
    Alibaba (Qwen)
    Score 92.8Q 92.8In $0.36/M
  2. 2
    Qwen3 Coder 480B A35B
    Alibaba (Qwen)
    Score 86.3Q 86.3In $0.30/M
  3. 3
    Kimi K3
    Moonshot (Kimi)
    Score 86.2Q 86.2In $3.00/M
  4. 4
    GLM-4.6
    Zhipu AI (GLM)
    Score 84.3Q 84.3In $0.60/M
  5. 5
    DeepSeek R1
    DeepSeek
    Score 83.0Q 83.0In $0.50/M
  6. 6
    MiniMax M2.5
    MiniMax
    Score 82.8Q 82.8In $0.30/M
  7. 7
    GLM 5.3
    Zhipu AI (GLM)
    Score 82.3Q 82.3In $1.40/M
  8. 8
    GLM 5
    Zhipu AI (GLM)
    Score 79.4Q 79.4In $1.00/M
  9. 9
    DeepSeek V3 (Thinking)
    DeepSeek
    Score 78.8Q 78.8In $0.29/M
  10. 10
    GLM 4.5
    Zhipu AI (GLM)
    Score 78.6Q 78.6In $0.60/M
  11. 11
    DeepSeek V4.1 Flash
    DeepSeek
    Score 75.6Q 75.6In $0.02/M
  12. 12
    DeepSeek V4 Pro 0813
    DeepSeek
    Score 75.1Q 75.1In $0.66/M
  13. 13
    Muse Spark Open
    Meta
    Score 74.1Q 74.1In $1.25/M
  14. 14
    MiniMax M2
    MiniMax
    Score 74.0Q 74.0In $0.26/M
  15. 15
    DeepSeek V4 Pro 0423
    DeepSeek
    Score 73.7Q 73.7In $0.21/M
Open interactive leaderboard Build custom weightsHome