Deepseek-V4-Pro-0813 Coding Benchmark
How Deepseek-V4-Pro-0813 ranks on the LLM Coding Leaderboard, shown next to the three models directly above and below it, with the top three models included for reference. Deepseek-V4-Pro-0813 was tested at 2 effort levels, each ranked separately.
Last evaluated: 1 week ago (November 14, 2026), with OpenCode.
| # | Model | Total points (max 90) |
Avg cost per prompt |
Avg time per prompt |
Video | Tested with | Points per project | Consistency penalty | ||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
Laravel Code Quality
(max 20) |
React-TS Code Quality
(max 20) |
Bug Finding Test
(max 10) |
Python Code Quality
(max 20) |
Edge-case Testing (4 projects, max 20) |
||||||||
| 1 | Opus 5.5 (High) | 84.95 | $0.83 | 03:42 |
Claude Code |
18.68 | 19 | 9.1 | 18.17 | 20 | 0 | |
| 2 | Fable 5.2 (Medium) | 84.7 | $1.42 | 03:03 |
Claude Code |
19.13 | 19.5 | 9.52 | 17.8 | 18.75 | 0 | |
| 3 | GPT-6-Astra (High) | 84.25 | $0.96 | 04:34 |
Codex CLI |
18.6 | 18.17 | 9.03 | 18.47 | 20 | −0.02 | |
| ⋯ 41 places not shown ⋯ | ||||||||||||
| 45 | GPT-6.1-Terra (Medium) | 68.22 | $0.04 | 00:31 |
Codex CLI |
16.28 | 16.17 | 7.43 | 15.67 | 14.45 | −1.78 | |
| 46 | GPT-6-Luna (High) | 67.33 | $0.01 | 03:51 |
Codex CLI |
16.9 | 15.83 | 5.35 | 15.53 | 15 | −1.28 | |
| 47 | Qwen 3.8 Flash (Xhigh) | 67.31 | $0.04 | 08:38 |
OpenCode |
16.35 | 17.67 | 5.7 | 16.33 | 13.2 | −1.94 | |
| 48 | Deepseek-V4-Pro-0813 (Max) | 66.9 | $0.05 | 09:06 |
OpenCode |
17.3 | 16.5 | 4.62 | 17.83 | 11.65 | −1.00 | |
| 49 | Deepseek-V4-Pro-0813 (High) | 65.07 | $0.04 | 07:15 |
OpenCode |
17.25 | 16.33 | 3.47 | 17.3 | 12.05 | −1.33 | |
| 50 | Mistral Large 4 (High) | 65.01 | $0.18 | 05:30 |
OpenCode |
17.78 | 18.5 | 4.14 | 15.03 | 12.4 | −2.84 | |
| 51 | Muse Ember 1.0 (Max) | 64.52 | $0.03 | 02:16 |
OpenCode |
15.95 | 16.67 | 5.31 | 14.33 | 14.7 | −2.44 | |
| 52 | MiMo-V2.6 Flash (High) | 64.39 | $0.04 | 17:56 |
OpenCode |
17.83 | 17.5 | 7.46 | 14.1 | 10.15 | −2.65 | |
Videos about Deepseek-V4-Pro-0813
I Tried NEW Deepseek-v4-Pro-0813: Better Than v4-Flash?
How to read this
- Each score measures a model-and-harness configuration, not the model in isolation.
- Prices are calculated with API costs. Models tested on a subscription plan show N/A.
- Full methodology and scoring formulas are explained in this article.
See every tested model side by side on the full LLM Coding Leaderboard.