Start
SquadOS
Language
EN
Português English
Theme
Log in

Compare AI models on accuracy, cost and time

36 combinations of model, agent tool and reasoning level ran the same task (video editing: remove silence) three times each.

How we measured

MakersOpenAIXiaomiAnthropicDeepSeekZ.ai

Agent toolsCodexSquadOS StudioClaude Code

Accuracy

Task criteria met, averaged over three runs.

  1. MiMo-V2.6-Flash on 89%
  2. GPT-6.1 Sol high Codex 89%
  3. MiMo-V2.6-Pro on 85%
  4. GPT-6.1 Sol extra high Codex 85%
  5. GPT-6.1 Sol max Codex 85%
  6. GPT-6.1 Sol low Codex 81%
  7. GPT-6.1 Sol medium Codex 81%
  8. DeepSeek V4.1 Flash max 81%
  9. Claude Sonnet 5.5 max Claude Code 81%
  10. GLM 5.3 max 81%
  11. DeepSeek V4.1 Flash high 78%
  12. Claude Opus 5.5 max Claude Code 78%
  13. GPT-6 Luna max Codex 78%
  14. Claude Opus 5.5 extra high Claude Code 78%
  15. Claude Opus 5.5 high Claude Code 78%
  16. DeepSeek V4.1 Flash low 74%
  17. Claude Sonnet 5.5 extra high Claude Code 74%
  18. MiMo-V2.6-Pro off 70%
  19. DeepSeek V4.1 Flash no reasoning 70%
  20. Claude Sonnet 5.5 high Claude Code 70%
  21. Claude Opus 5.5 medium Claude Code 70%
  22. GLM 5.3 Flash high 67%
  23. GLM 5.3 Flash low 67%
  24. Claude Haiku 5.5 max Claude Code 67%
  25. GPT-6 Luna extra high Codex 67%
  26. GPT-6 Luna low Codex 67%
  27. GLM 5.3 low 63%
  28. GPT-6 Luna high Codex 63%
  29. Claude Haiku 5.5 medium Claude Code 63%
  30. Claude Haiku 5.5 low Claude Code 59%
  31. Claude Opus 5.5 low Claude Code 59%
  32. Claude Haiku 5.5 high Claude Code 56%
  33. Claude Sonnet 5.5 medium Claude Code 52%
  34. Claude Haiku 5.5 extra high Claude Code 52%
  35. GPT-6 Luna medium Codex 48%
  36. Claude Sonnet 5.5 low Claude Code 44%

Cost per run

Estimated API cost in US dollars, averaged over three runs.

  1. GPT-6 Luna medium Codex $0.010
  2. GPT-6 Luna high Codex $0.010
  3. GPT-6 Luna low Codex $0.012
  4. Claude Haiku 5.5 low Claude Code $0.017
  5. Claude Haiku 5.5 medium Claude Code $0.019
  6. GPT-6 Luna max Codex $0.022
  7. Claude Haiku 5.5 high Claude Code $0.027
  8. GPT-6 Luna extra high Codex $0.031
  9. DeepSeek V4.1 Flash low $0.051
  10. DeepSeek V4.1 Flash high $0.076
  11. MiMo-V2.6-Flash on $0.082
  12. GLM 5.3 Flash low $0.087
  13. Claude Haiku 5.5 extra high Claude Code $0.092
  14. Claude Sonnet 5.5 low Claude Code $0.095
  15. GLM 5.3 low $0.099
  16. DeepSeek V4.1 Flash max $0.11
  17. GLM 5.3 Flash high $0.12
  18. MiMo-V2.6-Pro on $0.12
  19. Claude Sonnet 5.5 medium Claude Code $0.14
  20. DeepSeek V4.1 Flash no reasoning $0.15
  21. MiMo-V2.6-Pro off $0.22
  22. Claude Sonnet 5.5 high Claude Code $0.22
  23. GPT-6.1 Sol low Codex $0.23
  24. GPT-6.1 Sol medium Codex $0.26
  25. GPT-6.1 Sol extra high Codex $0.26
  26. GPT-6.1 Sol high Codex $0.28
  27. Claude Opus 5.5 low Claude Code $0.28
  28. GPT-6.1 Sol max Codex $0.33
  29. Claude Opus 5.5 medium Claude Code $0.38
  30. GLM 5.3 max $0.39
  31. Claude Sonnet 5.5 extra high Claude Code $0.48
  32. Claude Haiku 5.5 max Claude Code $0.50
  33. Claude Opus 5.5 high Claude Code $0.54
  34. Claude Opus 5.5 extra high Claude Code $1.37
  35. Claude Sonnet 5.5 max Claude Code $1.75
  36. Claude Opus 5.5 max Claude Code $4.86

Tokens per run

Input and output combined, including cache reads, averaged over three runs.

  1. Claude Sonnet 5.5 low Claude Code 124.7K
  2. Claude Sonnet 5.5 medium Claude Code 234K
  3. Claude Opus 5.5 low Claude Code 271.5K
  4. Claude Opus 5.5 medium Claude Code 348.5K
  5. GPT-6 Luna medium Codex 369.4K
  6. GPT-6 Luna high Codex 430.7K
  7. Claude Sonnet 5.5 high Claude Code 469.2K
  8. Claude Haiku 5.5 low Claude Code 492.5K
  9. Claude Opus 5.5 high Claude Code 565.3K
  10. GPT-6 Luna low Codex 594.4K
  11. Claude Haiku 5.5 medium Claude Code 642.3K
  12. GPT-6 Luna max Codex 720.1K
  13. GPT-6.1 Sol extra high Codex 796.3K
  14. GPT-6.1 Sol high Codex 880.3K
  15. GPT-6.1 Sol medium Codex 918.6K
  16. Claude Haiku 5.5 high Claude Code 937.2K
  17. GPT-6.1 Sol low Codex 960.1K
  18. GLM 5.3 low 984K
  19. GPT-6.1 Sol max Codex 999.3K
  20. Claude Sonnet 5.5 extra high Claude Code 1M
  21. GPT-6 Luna extra high Codex 1.1M
  22. GLM 5.3 Flash low 1.3M
  23. Claude Opus 5.5 extra high Claude Code 1.4M
  24. DeepSeek V4.1 Flash low 1.6M
  25. MiMo-V2.6-Pro on 2.1M
  26. Claude Haiku 5.5 extra high Claude Code 2.3M
  27. GLM 5.3 Flash high 2.5M
  28. DeepSeek V4.1 Flash high 2.7M
  29. Claude Sonnet 5.5 max Claude Code 3.7M
  30. GLM 5.3 max 4M
  31. MiMo-V2.6-Pro off 5.1M
  32. Claude Haiku 5.5 max Claude Code 5.3M
  33. DeepSeek V4.1 Flash max 5.8M
  34. Claude Opus 5.5 max Claude Code 6.3M
  35. MiMo-V2.6-Flash on 6.4M
  36. DeepSeek V4.1 Flash no reasoning 7.6M

Accuracy × cost

Dashed lines mark the median of each axis.

Accuracy
Cost per run, log scale

Accuracy × time

Dashed lines mark the median of each axis.

Accuracy
Time per run (median), log scale