IFEvalCode
Models write code in eight programming languages under explicit constraints. The evaluator checks functional correctness and adherence to instructions separately.
- Made by
- Jian Yang and colleagues
- Released
- July 2025
- Kind
- Function-level problems
- Level
- University
- Scored
- Checked automatically
- Answers
- Code
- Tasks
- All public
- Measures
- Capability
- Interaction
- Single turn
- Refresh
- One fixed set
- Run by
- Anyone
- Strongest model tested
- Claude 3.7 Sonnet #119 on the capability index
- Last model added
- April 2025
- Scores refreshed
- November 20th, 2026
Deepseek-R1 (snapshot unspecified) tops the leaderboard with 43.1% Claude 3.7 Sonnet leads with 38.2% among models released by April 2025
Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. The paper reports separate averages by prompt language. The headline is the plain mean of English and Chinese code correctness. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. The best model we track is Claude 3.7 Sonnet, at 38.2%. This view shows only models we track, each at its best.
The newest model with a score here came out in April 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 43.1 Deepseek-R1 (snapshot unspecified)
- 41.2 ControlledCoder-32B
- 38.2 Claude 3.7 Sonnet
- 36.6 GPT-4.1
- 34.5 Deepseek-V3 (snapshot unspecified)
- 34.4 GPT-4.1 mini
- 32.9 GPT-4o (snapshot unspecified)
- 32.5 Claude 3.5 Sonnet
- 31.3
Qwen3-30B-A3B
- 30.6
Qwen3-32B
- 30.4
Qwen3-235B-A22B
- 29.8
Qwen2.5-Coder-32B
- 29.3 Qwen2.5-Coder-14B
- 28.6 GPT-4o-mini (snapshot unspecified)
- 27.9
Llama 3.1-70B
- 27.0
Qwen3-14B
- 26.8
DeepSeek Coder 33B
- 26.6 Yi-Coder-9B
- 25.5
Qwen2.5-Coder (7B)
- 25.4 grok-3-mini-fast
- 22.3 OpenCoder-8B
- 21.8
DeepSeek Coder 6.7B
- 20.2 CodeQwen1.5-7B
- 19.6 Grok 3
- 17.9 OpenCoder-1.5B
- 17.6
Llama 3.1-8B
- 17.5 Granite-Coder-34B
- 17.5
Qwen3-8B
- 17.4 Qwen2.5-Coder-3B
- 16.4 Granite-Coder-8B
- 16.3 Yi-Coder-1.5B
- 16.2 Granite-Coder-20B
- 15.1
Qwen3-4B
- 12.7
Qwen2.5-Coder (1.5B)
- 12.6 Granite-Coder-3B
- 12.3 CodeLlama-13B
- 11.8
Qwen3-1.7B
- 11.3 CodeLlama-34B
- 10.5
DeepSeek Coder 1.3B
- 8.65 Qwen2.5-Coder-0.5B
- 7.35 CodeLlama-7B
- 5.40
Qwen3-0.6B
- 1 Deepseek-R1 (snapshot unspecified) 43.1%
- 2 ControlledCoder-32B 41.2%
- 3 Claude 3.7 Sonnet 38.2%
- 4 GPT-4.1 36.6%
- 5 Deepseek-V3 (snapshot unspecified) 34.5%
- 6 GPT-4.1 mini 34.4%
- 7 GPT-4o (snapshot unspecified) 32.9%
- 8 Claude 3.5 Sonnet 32.5%
- 9
Qwen3-30B-A3B Non-thinking 31.3%
- 10
Qwen3-32B Thinking 30.6%
- 11
Qwen3-235B-A22B Thinking 30.4%
- 12
Qwen2.5-Coder-32B 29.8%
- 13 Qwen2.5-Coder-14B 29.3%
- 14 GPT-4o-mini (snapshot unspecified) 28.6%
- 15
Llama 3.1-70B 27.9%
- 16
Qwen3-14B Thinking 27%
- 17
DeepSeek Coder 33B 26.8%
- 18 Yi-Coder-9B 26.6%
- 19
Qwen2.5-Coder (7B) 25.5%
- 20 grok-3-mini-fast 25.4%
- 21 OpenCoder-8B 22.3%
- 22
DeepSeek Coder 6.7B 21.8%
- 23 CodeQwen1.5-7B 20.2%
- 24 Grok 3 19.6%
- 25 OpenCoder-1.5B 17.9%
- 26
Llama 3.1-8B 17.6%
- 27 Granite-Coder-34B 17.5%
- 28
Qwen3-8B Thinking 17.5%
- 29 Qwen2.5-Coder-3B 17.4%
- 30 Granite-Coder-8B 16.4%
- 31 Yi-Coder-1.5B 16.3%
- 32 Granite-Coder-20B 16.2%
- 33
Qwen3-4B Non-thinking 15.1%
- 34
Qwen2.5-Coder (1.5B) 12.7%
- 35 Granite-Coder-3B 12.6%
- 36 CodeLlama-13B 12.3%
- 37
Qwen3-1.7B Non-thinking 11.8%
- 38 CodeLlama-34B 11.3%
- 39
DeepSeek Coder 1.3B 10.5%
- 40 Qwen2.5-Coder-0.5B 8.65%
- 41 CodeLlama-7B 7.35%
- 42
Qwen3-0.6B Non-thinking 5.40%
Source: The IFEvalCode paper
- 38.2 Claude 3.7 Sonnet
- 36.6 GPT-4.1
- 34.4 GPT-4.1 mini
- 32.5 Claude 3.5 Sonnet
- 31.3
Qwen3-30B-A3B
- 30.6
Qwen3-32B
- 30.4
Qwen3-235B-A22B
- 29.8
Qwen2.5-Coder-32B
- 29.3 Qwen2.5-Coder-14B
- 27.9
Llama 3.1-70B
- 27.0
Qwen3-14B
- 26.8
DeepSeek Coder 33B
- 25.5
Qwen2.5-Coder (7B)
- 21.8
DeepSeek Coder 6.7B
- 20.2 CodeQwen1.5-7B
- 19.6 Grok 3
- 17.6
Llama 3.1-8B
- 17.5
Qwen3-8B
- 17.4 Qwen2.5-Coder-3B
- 15.1
Qwen3-4B
- 12.7
Qwen2.5-Coder (1.5B)
- 11.8
Qwen3-1.7B
- 10.5
DeepSeek Coder 1.3B
- 8.65 Qwen2.5-Coder-0.5B
- 5.40
Qwen3-0.6B
- 1 Claude 3.7 Sonnet 38.2%
- 2 GPT-4.1 36.6%
- 3 GPT-4.1 mini 34.4%
- 4 Claude 3.5 Sonnet 32.5%
- 5
Qwen3-30B-A3B Non-thinking 31.3%
- 6
Qwen3-32B Thinking 30.6%
- 7
Qwen3-235B-A22B Thinking 30.4%
- 8
Qwen2.5-Coder-32B 29.8%
- 9 Qwen2.5-Coder-14B 29.3%
- 10
Llama 3.1-70B 27.9%
- 11
Qwen3-14B Thinking 27%
- 12
DeepSeek Coder 33B 26.8%
- 13
Qwen2.5-Coder (7B) 25.5%
- 14
DeepSeek Coder 6.7B 21.8%
- 15 CodeQwen1.5-7B 20.2%
- 16 Grok 3 19.6%
- 17
Llama 3.1-8B 17.6%
- 18
Qwen3-8B Thinking 17.5%
- 19 Qwen2.5-Coder-3B 17.4%
- 20
Qwen3-4B Non-thinking 15.1%
- 21
Qwen2.5-Coder (1.5B) 12.7%
- 22
Qwen3-1.7B Non-thinking 11.8%
- 23
DeepSeek Coder 1.3B 10.5%
- 24 Qwen2.5-Coder-0.5B 8.65%
- 25
Qwen3-0.6B Non-thinking 5.40%
Source: The IFEvalCode paper
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Deepseek-R1 (snapshot unspecified) | 43.1% | Official leaderboard | ||
| 2 | ControlledCoder-32B | 41.2% | Official leaderboard | ||
| 3 | Claude 3.7 Sonnet | 38.2% | Feb 2025 | Official leaderboard | |
| 4 | GPT-4.1 | 36.6% | Apr 2025 | Official leaderboard | |
| 5 | Deepseek-V3 (snapshot unspecified) | 34.5% | Official leaderboard | ||
| 6 | GPT-4.1 mini | 34.4% | Apr 2025 | Official leaderboard | |
| 7 | GPT-4o (snapshot unspecified) | 32.9% | Official leaderboard | ||
| 8 | Claude 3.5 Sonnet | 32.5% | Jun 2024 | Official leaderboard | |
| 9 | | 31.3% | Non-thinking | Apr 2025 | Official leaderboard |
| 10 | | 30.6% | Thinking | Apr 2025 | Official leaderboard |
| 11 | | 30.4% | Thinking | Apr 2025 | Official leaderboard |
| 12 | | 29.8% | Sep 2024 | Official leaderboard | |
| 13 | Qwen2.5-Coder-14B | 29.3% | Sep 2024 | Official leaderboard | |
| 14 | GPT-4o-mini (snapshot unspecified) | 28.6% | Official leaderboard | ||
| 15 | | 27.9% | Jul 2024 | Official leaderboard | |
| 16 | | 27% | Thinking | Apr 2025 | Official leaderboard |
| 17 | | 26.8% | Nov 2023 | Official leaderboard | |
| 18 | Yi-Coder-9B | 26.6% | Official leaderboard | ||
| 19 | | 25.5% | Sep 2024 | Official leaderboard | |
| 20 | grok-3-mini-fast | 25.4% | Official leaderboard | ||
| 21 | OpenCoder-8B | 22.3% | Official leaderboard | ||
| 22 | | 21.8% | Nov 2023 | Official leaderboard | |
| 23 | CodeQwen1.5-7B | 20.2% | Apr 2024 | Official leaderboard | |
| 24 | Grok 3 | 19.6% | Apr 2025 | Official leaderboard | |
| 25 | OpenCoder-1.5B | 17.9% | Official leaderboard | ||
| 26 | | 17.6% | Jul 2024 | Official leaderboard | |
| 27 | Granite-Coder-34B | 17.5% | Official leaderboard | ||
| 28 | | 17.5% | Thinking | Apr 2025 | Official leaderboard |
| 29 | Qwen2.5-Coder-3B | 17.4% | Sep 2024 | Official leaderboard | |
| 30 | Granite-Coder-8B | 16.4% | Official leaderboard | ||
| 31 | Yi-Coder-1.5B | 16.3% | Official leaderboard | ||
| 32 | Granite-Coder-20B | 16.2% | Official leaderboard | ||
| 33 | | 15.1% | Non-thinking | Apr 2025 | Official leaderboard |
| 34 | | 12.7% | Sep 2024 | Official leaderboard | |
| 35 | Granite-Coder-3B | 12.6% | Official leaderboard | ||
| 36 | CodeLlama-13B | 12.3% | Official leaderboard | ||
| 37 | | 11.8% | Non-thinking | Apr 2025 | Official leaderboard |
| 38 | CodeLlama-34B | 11.3% | Official leaderboard | ||
| 39 | | 10.5% | Nov 2023 | Official leaderboard | |
| 40 | Qwen2.5-Coder-0.5B | 8.65% | Sep 2024 | Official leaderboard | |
| 41 | CodeLlama-7B | 7.35% | Official leaderboard | ||
| 42 | | 5.40% | Non-thinking | Apr 2025 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude 3.7 Sonnet | 38.2% | Feb 2025 | Official leaderboard | |
| 2 | GPT-4.1 | 36.6% | Apr 2025 | Official leaderboard | |
| 3 | GPT-4.1 mini | 34.4% | Apr 2025 | Official leaderboard | |
| 4 | Claude 3.5 Sonnet | 32.5% | Jun 2024 | Official leaderboard | |
| 5 | | 31.3% | Non-thinking | Apr 2025 | Official leaderboard |
| 6 | | 30.6% | Thinking | Apr 2025 | Official leaderboard |
| 7 | | 30.4% | Thinking | Apr 2025 | Official leaderboard |
| 8 | | 29.8% | Sep 2024 | Official leaderboard | |
| 9 | Qwen2.5-Coder-14B | 29.3% | Sep 2024 | Official leaderboard | |
| 10 | | 27.9% | Jul 2024 | Official leaderboard | |
| 11 | | 27% | Thinking | Apr 2025 | Official leaderboard |
| 12 | | 26.8% | Nov 2023 | Official leaderboard | |
| 13 | | 25.5% | Sep 2024 | Official leaderboard | |
| 14 | | 21.8% | Nov 2023 | Official leaderboard | |
| 15 | CodeQwen1.5-7B | 20.2% | Apr 2024 | Official leaderboard | |
| 16 | Grok 3 | 19.6% | Apr 2025 | Official leaderboard | |
| 17 | | 17.6% | Jul 2024 | Official leaderboard | |
| 18 | | 17.5% | Thinking | Apr 2025 | Official leaderboard |
| 19 | Qwen2.5-Coder-3B | 17.4% | Sep 2024 | Official leaderboard | |
| 20 | | 15.1% | Non-thinking | Apr 2025 | Official leaderboard |
| 21 | | 12.7% | Sep 2024 | Official leaderboard | |
| 22 | | 11.8% | Non-thinking | Apr 2025 | Official leaderboard |
| 23 | | 10.5% | Nov 2023 | Official leaderboard | |
| 24 | Qwen2.5-Coder-0.5B | 8.65% | Sep 2024 | Official leaderboard | |
| 25 | | 5.40% | Non-thinking | Apr 2025 | Official leaderboard |
Rankings by part
We read 36 parts from the benchmark’s paper, each ranking models on one part of the test, such as English Python Correctness, English Python Instruction adherence and English Java Correctness. Claude 3.7 Sonnet leads 8 of them, and 11 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
English Python Correctness
25 models we track have a score here.
- 1
Qwen3-30B-A3B ↑4 54.3%
- 2
Qwen3-14B ↑9 51.4%
- 3 GPT-4.1 ↓1 50.5%
- 4
Qwen3-235B-A22B ↑3 50.5%
- 5
Qwen3-32B ↑1 49.5%
- 6 Claude 3.5 Sonnet ↓2 48.6%
- 7 Qwen2.5-Coder-14B ↑2 48.6%
- 8 Claude 3.7 Sonnet ↓7 47.6%
- 9 GPT-4.1 mini ↓6 46.7%
- 10
Llama 3.1-70B 42.9%
English Python Instruction adherence
25 models we track have a score here.
- 1 GPT-4.1 mini ↑2 34.3%
- 2
Llama 3.1-70B ↑8 33.3%
- 3
Qwen2.5-Coder-32B ↑5 33.3%
- 4
Qwen3-14B ↑7 33.3%
- 5
Qwen3-235B-A22B ↑2 33.3%
- 6
Qwen3-30B-A3B ↓1 33.3%
- 7
Qwen3-32B ↓1 33.3%
- 8
Qwen3-8B ↑10 33.3%
- 9 Claude 3.7 Sonnet ↓8 31.4%
- 10 GPT-4.1 ↓8 31.4%
English Java Correctness
25 models we track have a score here.
- 1 Claude 3.7 Sonnet 37.3%
- 2 GPT-4.1 36.3%
- 3 GPT-4.1 mini 35.3%
- 4
DeepSeek Coder 33B ↑8 33.3%
- 5 Claude 3.5 Sonnet ↓1 33.3%
- 6
Qwen3-30B-A3B ↓1 30.4%
- 7
Qwen3-235B-A22B 28.4%
- 8
DeepSeek Coder 6.7B ↑6 27.5%
- 9 Qwen2.5-Coder-14B 27.5%
- 10
Qwen2.5-Coder (7B) ↑3 25.5%
English Java Instruction adherence
25 models we track have a score here.
- 1
Qwen3-14B ↑10 15.7%
- 2 Grok 3 ↑14 13.7%
- 3 GPT-4.1 ↓1 12.7%
- 4
Qwen3-30B-A3B ↑1 11.8%
- 5 Claude 3.5 Sonnet ↓1 10.8%
- 6
Qwen3-235B-A22B ↑1 10.8%
- 7
Qwen3-8B ↑11 9.80%
- 8
Qwen3-32B ↓2 8.80%
- 9
DeepSeek Coder 6.7B ↑5 7.80%
- 10 Qwen2.5-Coder-14B ↓1 7.80%
English Cpp Correctness
25 models we track have a score here.
- 1
Qwen3-30B-A3B ↑4 31%
- 2 Claude 3.7 Sonnet ↓1 29%
- 3 GPT-4.1 ↓1 27%
- 4 GPT-4.1 mini ↓1 23%
- 5
DeepSeek Coder 33B ↑7 22%
- 6 Qwen2.5-Coder-14B ↑3 22%
- 7
DeepSeek Coder 6.7B ↑7 20%
- 8
Llama 3.1-70B ↑2 20%
- 9
Qwen2.5-Coder-32B ↓1 20%
- 10
Qwen3-14B ↑1 20%
English Cpp Instruction adherence
25 models we track have a score here.
- 1
Qwen3-235B-A22B ↑6 12%
- 2 Claude 3.5 Sonnet ↑2 9.00%
- 3
Qwen3-32B ↑3 7.00%
- 4
Llama 3.1-70B ↑6 6.00%
- 5
Qwen2.5-Coder-32B ↑3 6.00%
- 6 Claude 3.7 Sonnet ↓5 6.00%
- 7 Grok 3 ↑9 6.00%
- 8 GPT-4.1 ↓6 6.00%
- 9
Qwen3-30B-A3B ↓4 6.00%
- 10
Qwen3-4B ↑10 6.00%
English C-sharp Correctness
25 models we track have a score here.
- 1 Qwen2.5-Coder-14B ↑8 18.4%
- 2 Claude 3.7 Sonnet ↓1 16.5%
- 3 Claude 3.5 Sonnet ↑1 12.6%
- 4 Qwen2.5-Coder-3B ↑15 11.7%
- 5
Qwen2.5-Coder (7B) ↑8 11.7%
- 6
Llama 3.1-8B ↑11 10.7%
- 7
Qwen3-32B ↓1 10.7%
- 8 Qwen2.5-Coder-0.5B ↑16 9.70%
- 9
Llama 3.1-70B ↑1 8.70%
- 10
Qwen2.5-Coder-32B ↓2 8.70%
English C-sharp Instruction adherence
25 models we track have a score here.
- 1
Qwen3-235B-A22B ↑6 6.80%
- 2
Qwen3-8B ↑16 5.80%
- 3
Llama 3.1-70B ↑7 4.90%
- 4
Qwen2.5-Coder-32B ↑4 4.90%
- 5
Qwen3-14B ↑6 4.90%
- 6
Qwen3-30B-A3B ↓1 4.90%
- 7 Claude 3.7 Sonnet ↓6 3.90%
- 8 Grok 3 ↑8 3.90%
- 9 GPT-4.1 ↓7 3.90%
- 10
Qwen3-4B ↑10 3.90%
English Typescript Correctness
25 models we track have a score here.
- 1
Qwen3-30B-A3B ↑4 32%
- 2 GPT-4.1 30%
- 3
Qwen3-14B ↑8 30%
- 4
Llama 3.1-70B ↑6 29%
- 5
Qwen2.5-Coder-32B ↑3 29%
- 6 Claude 3.7 Sonnet ↓5 29%
- 7 GPT-4.1 mini ↓4 27%
- 8
DeepSeek Coder 33B ↑4 24%
- 9 Claude 3.5 Sonnet ↓5 24%
- 10
Qwen3-235B-A22B ↓3 21%
English Typescript Instruction adherence
25 models we track have a score here.
- 1
Llama 3.1-70B ↑9 30%
- 2
Qwen2.5-Coder-32B ↑6 30%
- 3
Qwen3-30B-A3B ↑2 30%
- 4
Qwen3-32B ↑2 29%
- 5 Claude 3.7 Sonnet ↓4 28%
- 6 GPT-4.1 ↓4 27%
- 7 Claude 3.5 Sonnet ↓3 25%
- 8 Qwen2.5-Coder-14B ↑1 25%
- 9
Qwen3-4B ↑11 25%
- 10
Qwen3-14B ↑1 24%
English Javascript Correctness
25 models we track have a score here.
- 1 GPT-4.1 ↑1 58%
- 2 GPT-4.1 mini ↑1 57%
- 3 Claude 3.7 Sonnet ↓2 54%
- 4 Claude 3.5 Sonnet 48%
- 5
Qwen3-14B ↑6 48%
- 6
Qwen3-30B-A3B ↓1 48%
- 7
Llama 3.1-70B ↑3 47%
- 8
Qwen2.5-Coder-32B 47%
- 9 Grok 3 ↑7 44%
- 10
Qwen3-235B-A22B ↓3 43%
English Javascript Instruction adherence
25 models we track have a score here.
- 1 Claude 3.7 Sonnet 34%
- 2
Qwen3-32B ↑4 34%
- 3
Qwen3-14B ↑8 33%
- 4
Qwen3-30B-A3B ↑1 33%
- 5 Claude 3.5 Sonnet ↓1 32%
- 6 Qwen2.5-Coder-14B ↑3 31%
- 7
Llama 3.1-70B ↑3 30%
- 8
Qwen2.5-Coder-32B 30%
- 9
Qwen2.5-Coder (7B) ↑4 30%
- 10 GPT-4.1 mini ↓7 30%
English Php Correctness
25 models we track have a score here.
- 1 Claude 3.7 Sonnet 59%
- 2 GPT-4.1 58%
- 3 GPT-4.1 mini 50%
- 4
Qwen3-235B-A22B ↑3 47%
- 5
Qwen3-30B-A3B 46%
- 6
Qwen2.5-Coder (7B) ↑7 43%
- 7 Grok 3 ↑9 43%
- 8
Qwen3-14B ↑3 43%
- 9
Llama 3.1-70B ↑1 42%
- 10
Qwen2.5-Coder-32B ↓2 42%
English Php Instruction adherence
25 models we track have a score here.
- 1 GPT-4.1 mini ↑2 29%
- 2
Qwen3-32B ↑4 28%
- 3 Claude 3.7 Sonnet ↓2 27%
- 4 GPT-4.1 ↓2 26%
- 5 Grok 3 ↑11 24%
- 6 Claude 3.5 Sonnet ↓2 20%
- 7
Qwen3-14B ↑4 20%
- 8
Qwen3-8B ↑10 19%
- 9
Llama 3.1-70B ↑1 16%
- 10
Qwen2.5-Coder-32B ↓2 16%
English Shell Correctness
25 models we track have a score here.
- 1 GPT-4.1 ↑1 40%
- 2 GPT-4.1 mini ↑1 38%
- 3 Claude 3.7 Sonnet ↓2 36%
- 4
Qwen3-30B-A3B ↑1 36%
- 5
Qwen3-235B-A22B ↑2 35%
- 6 Claude 3.5 Sonnet ↓2 28%
- 7
Llama 3.1-70B ↑3 27%
- 8
Qwen2.5-Coder-32B 27%
- 9
Qwen3-14B ↑2 24%
- 10
Qwen3-32B ↓4 22%
English Shell Instruction adherence
25 models we track have a score here.
- 1 GPT-4.1 mini ↑2 53%
- 2
Qwen3-14B ↑9 53%
- 3 GPT-4.1 ↓1 52%
- 4
Qwen3-32B ↑2 52%
- 5 Claude 3.7 Sonnet ↓4 43%
- 6
Llama 3.1-70B ↑4 39%
- 7
Qwen2.5-Coder-32B ↑1 39%
- 8
Qwen3-4B ↑12 38%
- 9 Qwen2.5-Coder-14B 35%
- 10
Qwen3-235B-A22B ↓3 35%
English Correctness
25 models we track have a score here.
- 1 Claude 3.7 Sonnet 38.5%
- 2 GPT-4.1 36.8%
- 3
Qwen3-30B-A3B ↑2 34.7%
- 4 GPT-4.1 mini ↓1 34.6%
- 5 Claude 3.5 Sonnet ↓1 31.9%
- 6
Llama 3.1-70B ↑4 29.9%
- 7
Qwen2.5-Coder-32B ↑1 29.9%
- 8
Qwen3-14B ↑3 29.6%
- 9
Qwen3-235B-A22B ↓2 29.1%
- 10 Qwen2.5-Coder-14B ↓1 28.6%
English Instruction adherence
25 models we track have a score here.
- 1
Qwen3-32B ↑5 24.2%
- 2 GPT-4.1 23%
- 3 GPT-4.1 mini 23%
- 4 Claude 3.7 Sonnet ↓3 22.6%
- 5
Qwen3-14B ↑6 21.9%
- 6
Llama 3.1-70B ↑4 20.7%
- 7
Qwen2.5-Coder-32B ↑1 20.7%
- 8 Claude 3.5 Sonnet ↓4 20%
- 9
Qwen3-235B-A22B ↓2 19.3%
- 10 Grok 3 ↑6 18.8%
Chinese Python Correctness
25 models we track have a score here.
- 1
Qwen3-32B ↑5 54.3%
- 2 GPT-4.1 51.4%
- 3 Claude 3.5 Sonnet ↑1 49.5%
- 4
Qwen3-14B ↑7 49.5%
- 5
Llama 3.1-70B ↑5 47.6%
- 6 Qwen2.5-Coder-14B ↑3 47.6%
- 7 Claude 3.7 Sonnet ↓6 47.6%
- 8 GPT-4.1 mini ↓5 46.7%
- 9
Qwen3-235B-A22B ↓2 46.7%
- 10
DeepSeek Coder 33B ↑2 44.8%
Chinese Python Instruction adherence
25 models we track have a score here.
- 1
Qwen3-32B ↑5 41%
- 2 GPT-4.1 39%
- 3
Qwen3-235B-A22B ↑4 38.1%
- 4 GPT-4.1 mini ↓1 36.2%
- 5
Qwen2.5-Coder-32B ↑3 34.3%
- 6 Claude 3.5 Sonnet ↓2 33.3%
- 7
Qwen3-30B-A3B ↓2 33.3%
- 8
Qwen3-14B ↑3 32.4%
- 9
Qwen3-8B ↑9 32.4%
- 10
Qwen3-4B ↑10 30.5%
Chinese Java Correctness
25 models we track have a score here.
- 1 Claude 3.5 Sonnet ↑3 38.2%
- 2 GPT-4.1 mini ↑1 38.2%
- 3 Claude 3.7 Sonnet ↓2 36.3%
- 4 GPT-4.1 ↓2 35.3%
- 5
Qwen3-32B ↑1 30.4%
- 6
Qwen2.5-Coder-32B ↑2 28.4%
- 7
Qwen3-30B-A3B ↓2 28.4%
- 8
DeepSeek Coder 33B ↑4 26.5%
- 9 Qwen2.5-Coder-14B 26.5%
- 10
Qwen3-235B-A22B ↓3 26.5%
Chinese Java Instruction adherence
25 models we track have a score here.
- 1 Claude 3.5 Sonnet ↑3 11.8%
- 2 GPT-4.1 11.8%
- 3
Qwen3-30B-A3B ↑2 11.8%
- 4 Grok 3 ↑12 9.80%
- 5 Qwen2.5-Coder-14B ↑4 8.80%
- 6
Qwen2.5-Coder-32B ↑2 8.80%
- 7
Qwen3-14B ↑4 8.80%
- 8
Qwen3-235B-A22B ↓1 8.80%
- 9 Claude 3.7 Sonnet ↓8 7.80%
- 10
Qwen3-32B ↓4 7.80%
Chinese Cpp Correctness
25 models we track have a score here.
- 1 Claude 3.7 Sonnet 28%
- 2
Qwen3-235B-A22B ↑5 26%
- 3 GPT-4.1 ↓1 25%
- 4
DeepSeek Coder 33B ↑8 24%
- 5
Qwen3-32B ↑1 24%
- 6 Claude 3.5 Sonnet ↓2 23%
- 7 Qwen2.5-Coder-14B ↑2 22%
- 8
Llama 3.1-70B ↑2 19%
- 9
Qwen2.5-Coder-32B ↓1 19%
- 10
Qwen3-30B-A3B ↓5 19%
Chinese Cpp Instruction adherence
25 models we track have a score here.
- 1
DeepSeek Coder 33B ↑11 7.00%
- 2 Claude 3.5 Sonnet ↑2 7.00%
- 3 Claude 3.7 Sonnet ↓2 7.00%
- 4 GPT-4.1 ↓2 7.00%
- 5
Qwen3-32B ↑1 7.00%
- 6
Qwen3-30B-A3B ↓1 6.00%
- 7
Qwen3-0.6B ↑18 6.00%
- 8
Qwen2.5-Coder (1.5B) ↑13 5.00%
- 9 Qwen2.5-Coder-3B ↑10 5.00%
- 10
Qwen2.5-Coder (7B) ↑3 5.00%
Chinese C-sharp Correctness
25 models we track have a score here.
- 1 Qwen2.5-Coder-14B ↑8 19.4%
- 2
Llama 3.1-8B ↑15 15.5%
- 3 Claude 3.5 Sonnet ↑1 14.6%
- 4
Qwen2.5-Coder-32B ↑4 12.6%
- 5 Claude 3.7 Sonnet ↓4 12.6%
- 6
Llama 3.1-70B ↑4 11.7%
- 7 Qwen2.5-Coder-3B ↑12 11.7%
- 8
Qwen3-14B ↑3 10.7%
- 9 CodeQwen1.5-7B ↑6 9.70%
- 10
Qwen3-32B ↓4 9.70%
Chinese C-sharp Instruction adherence
25 models we track have a score here.
- 1
Qwen2.5-Coder-32B ↑7 5.80%
- 2
Qwen3-30B-A3B ↑3 5.80%
- 3
Qwen3-14B ↑8 4.90%
- 4 Claude 3.7 Sonnet ↓3 3.90%
- 5
Qwen3-32B ↑1 3.90%
- 6
Qwen3-8B ↑12 3.90%
- 7
Qwen3-4B ↑13 3.90%
- 8 Claude 3.5 Sonnet ↓4 2.90%
- 9
Llama 3.1-70B ↑1 2.90%
- 10 Qwen2.5-Coder-14B ↓1 2.90%
Chinese Typescript Correctness
25 models we track have a score here.
- 1 GPT-4.1 ↑1 35%
- 2 GPT-4.1 mini ↑1 32%
- 3
Qwen3-235B-A22B ↑4 32%
- 4 Claude 3.7 Sonnet ↓3 30%
- 5
Qwen3-32B ↑1 27%
- 6
Qwen2.5-Coder-32B ↑2 25%
- 7 Qwen2.5-Coder-14B ↑2 24%
- 8
Qwen3-30B-A3B ↓3 24%
- 9
Qwen3-14B ↑2 23%
- 10
DeepSeek Coder 33B ↑2 21%
Chinese Typescript Instruction adherence
25 models we track have a score here.
- 1 Qwen2.5-Coder-14B ↑8 32%
- 2
Qwen3-14B ↑9 31%
- 3
Qwen3-30B-A3B ↑2 31%
- 4 Claude 3.7 Sonnet ↓3 30%
- 5 Claude 3.5 Sonnet ↓1 29%
- 6
Qwen2.5-Coder-32B ↑2 29%
- 7 GPT-4.1 mini ↓4 28%
- 8 GPT-4.1 ↓6 27%
- 9
Qwen3-4B ↑11 26%
- 10
Qwen3-32B ↓4 25%
Chinese Javascript Correctness
25 models we track have a score here.
- 1 Claude 3.7 Sonnet 51%
- 2 GPT-4.1 51%
- 3 GPT-4.1 mini 50%
- 4 Grok 3 ↑12 46%
- 5 Claude 3.5 Sonnet ↓1 45%
- 6
Qwen3-235B-A22B ↑1 44%
- 7
Qwen3-30B-A3B ↓2 42%
- 8
Qwen3-32B ↓2 42%
- 9 Qwen2.5-Coder-14B 41%
- 10
Qwen2.5-Coder-32B ↓2 41%
Chinese Javascript Instruction adherence
25 models we track have a score here.
- 1
Qwen2.5-Coder-32B ↑7 34%
- 2
Qwen3-14B ↑9 33%
- 3 Claude 3.5 Sonnet ↑1 32%
- 4
Qwen3-8B ↑14 32%
- 5
Qwen3-30B-A3B 31%
- 6
Qwen3-32B 31%
- 7
Qwen3-4B ↑13 31%
- 8 Qwen2.5-Coder-14B ↑1 30%
- 9 Claude 3.7 Sonnet ↓8 30%
- 10
Qwen3-235B-A22B ↓3 30%
Chinese Php Correctness
25 models we track have a score here.
- 1 Claude 3.7 Sonnet 61%
- 2 GPT-4.1 54%
- 3
Qwen3-32B ↑3 50%
- 4 Claude 3.5 Sonnet 47%
- 5
Qwen3-235B-A22B ↑2 47%
- 6 GPT-4.1 mini ↓3 46%
- 7
Qwen2.5-Coder-32B ↑1 44%
- 8
Qwen3-14B ↑3 40%
- 9
Qwen3-30B-A3B ↓4 40%
- 10
Llama 3.1-70B 39%
Chinese Php Instruction adherence
25 models we track have a score here.
- 1 GPT-4.1 ↑1 35%
- 2 GPT-4.1 mini ↑1 33%
- 3 Grok 3 ↑13 32%
- 4 Claude 3.5 Sonnet 29%
- 5
Qwen3-30B-A3B 29%
- 6
Qwen3-8B ↑12 27%
- 7
Qwen2.5-Coder-32B ↑1 24%
- 8
Qwen3-14B ↑3 23%
- 9
Qwen3-32B ↓3 23%
- 10 Claude 3.7 Sonnet ↓9 22%
Chinese Shell Correctness
25 models we track have a score here.
- 1 GPT-4.1 mini ↑2 43%
- 2 GPT-4.1 40%
- 3 Claude 3.7 Sonnet ↓2 37%
- 4
Qwen3-235B-A22B ↑3 36%
- 5
Qwen3-32B ↑1 30%
- 6 Claude 3.5 Sonnet ↓2 27%
- 7
Qwen2.5-Coder-32B ↑1 23%
- 8
Qwen3-14B ↑3 22%
- 9
Qwen3-30B-A3B ↓4 22%
- 10
Llama 3.1-70B 21%
Chinese Shell Instruction adherence
25 models we track have a score here.
- 1
Qwen3-14B ↑10 53%
- 2 GPT-4.1 mini ↑1 52%
- 3
Qwen3-32B ↑3 51%
- 4
Qwen3-30B-A3B ↑1 48%
- 5
Qwen3-4B ↑15 47%
- 6 GPT-4.1 ↓4 46%
- 7 Claude 3.7 Sonnet ↓6 44%
- 8
Llama 3.1-70B ↑2 40%
- 9
Qwen2.5-Coder-32B ↓1 38%
- 10
Qwen3-235B-A22B ↓3 38%
Chinese Correctness
25 models we track have a score here.
- 1 Claude 3.7 Sonnet 37.9%
- 2 GPT-4.1 36.4%
- 3 GPT-4.1 mini 34.2%
- 4 Claude 3.5 Sonnet 33.1%
- 5
Qwen3-32B ↑1 33.1%
- 6
Qwen3-235B-A22B ↑1 31.7%
- 7 Qwen2.5-Coder-14B ↑2 29.9%
- 8
Qwen2.5-Coder-32B 29.6%
- 9
Qwen3-30B-A3B ↓4 27.8%
- 10
DeepSeek Coder 33B ↑2 26.2%
Chinese Instruction adherence
25 models we track have a score here.
- 1 GPT-4.1 ↑1 24.7%
- 2
Qwen3-30B-A3B ↑3 24.4%
- 3 GPT-4.1 mini 23.7%
- 4
Qwen3-32B ↑2 23.7%
- 5
Qwen3-14B ↑6 23.6%
- 6 Claude 3.5 Sonnet ↓2 22.6%
- 7
Qwen2.5-Coder-32B ↑1 22%
- 8 Claude 3.7 Sonnet ↓7 21.6%
- 9
Qwen3-4B ↑11 21.2%
- 10
Qwen3-235B-A22B ↓3 20.9%
Scores at each effort level
Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.
Source: The IFEvalCode paper
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| Claude 3.7 Sonnet | 38.2% | ||||
| GPT-4.1 | 36.6% | ||||
| GPT-4.1 mini | 34.4% | ||||
| Claude 3.5 Sonnet | 32.5% | ||||
| Non-thinking | 31.3% | ||||
| Thinking | 30.6% | ||||
| Thinking | 30.4% | ||||
| 29.8% | |||||
| Qwen2.5-Coder-14B | 29.3% | ||||
| 27.9% | |||||
| Non-thinking | 27.8% | ||||
| Non-thinking | 27.6% | ||||
| Thinking | 27% | ||||
| 26.8% | |||||
| Thinking | 26.4% | ||||
| Non-thinking | 25.8% | ||||
| 25.5% | |||||
| 21.8% | |||||
| CodeQwen1.5-7B | 20.2% | ||||
| Grok 3 | 19.6% | ||||
| 17.6% | |||||
| Thinking | 17.5% | ||||
| Qwen2.5-Coder-3B | 17.4% | ||||
| Non-thinking | 16.7% | ||||
| Non-thinking | 15.1% | ||||
| Thinking | 14.1% | ||||
| 12.7% | |||||
| Non-thinking | 11.8% | ||||
| Thinking | 10.8% | ||||
| 10.5% | |||||
| Qwen2.5-Coder-0.5B | 8.65% | ||||
| Non-thinking | 5.40% | ||||
| Thinking | 4.30% |
Scores at each effort level
Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.
Source: The IFEvalCode paper
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| Claude 3.7 Sonnet | 38.2% | ||||
| GPT-4.1 | 36.6% | ||||
| GPT-4.1 mini | 34.4% | ||||
| Claude 3.5 Sonnet | 32.5% | ||||
| Non-thinking | 31.3% | ||||
| Thinking | 30.6% | ||||
| Thinking | 30.4% | ||||
| 29.8% | |||||
| Qwen2.5-Coder-14B | 29.3% | ||||
| 27.9% | |||||
| Non-thinking | 27.8% | ||||
| Non-thinking | 27.6% | ||||
| Thinking | 27% | ||||
| 26.8% | |||||
| Thinking | 26.4% | ||||
| Non-thinking | 25.8% | ||||
| 25.5% | |||||
| 21.8% | |||||
| CodeQwen1.5-7B | 20.2% | ||||
| Grok 3 | 19.6% | ||||
| 17.6% | |||||
| Thinking | 17.5% | ||||
| Qwen2.5-Coder-3B | 17.4% | ||||
| Non-thinking | 16.7% | ||||
| Non-thinking | 15.1% | ||||
| Thinking | 14.1% | ||||
| 12.7% | |||||
| Non-thinking | 11.8% | ||||
| Thinking | 10.8% | ||||
| 10.5% | |||||
| Qwen2.5-Coder-0.5B | 8.65% | ||||
| Non-thinking | 5.40% | ||||
| Thinking | 4.30% |
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: The IFEvalCode paper
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official IFEvalCode leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Coding
- SWE-bench Verified In use 500 human-validated GitHub issues from 12 Python repos. Agent writes a patch that must pass hidden tests. Claude Opus 4.7 83.5%
- LiveCodeBench v6 In use Competitive programming problems collected after model cutoffs from LeetCode, AtCoder, Codeforces. Scored pass@1 on hidden tests.
DeepSeek-V4-Pro 93.5%
- SWE-bench Multilingual In use 300 GitHub-issue tasks across nine programming languages in SWE-bench format. Patch graded by repository tests. Claude Fable 5.2 96.6%
- DeepSWE In use 113 original long-horizon engineering tasks across 91 repos and five languages, written from scratch. Hand-written verifiers accept alternative solutions. GPT-6 Astra 74.1%
- SciCode In use Scientist-curated research coding problems split into subproblems across natural sciences. Code checked against test cases.
Seed2.1 Pro 59.8%
- Codeforces rating In use Model solves recent Codeforces contest problems. An Elo-style rating is computed from simulated contest performance.
DeepSeek V4.1 Flash 3,471
Data audit
9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 19 months ago, in April 2025, for Qwen3-0.6B. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
- Pass Independent runs Scores from the benchmark’s paper (42 entries).
- Pass Models scored 25 models we track, plus 17 other entries on its leaderboard.
- Warning Newest score The newest came 19 months ago, in April 2025, for Qwen3-0.6B.
- Note Frontier models None of the top 10 has a score. The strongest tested is Claude 3.7 Sonnet (#119).
- Pass Headroom The best is 38.2%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on arxiv.org: 33 rows for models we track and 17 other entries at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.