Benchmark Almanac

AI Engineering Lab model creation

Qwen2.5-Coder-32B-Instruct 91.7% leads since April 2025
Not in model cards

Models generate Exudyn simulation code for 35 mechanical systems. The evaluator compares numerical outputs with reference simulations across ten parameter settings per system.

The authors publish results for 14 model configurations on 350 cases. Two configurations use different implementations of Llama 3.1 8B.

Released
April 2025
Kind
Scientific code
Level
University
Scored
Checked automatically
Answers
Code
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Code
Interaction
Single turn
Environment
Code sandbox
Items from
Generated
Run by
Anyone
Strongest model tested
Phi-4 #160 on the capability index
Last model added
January 2025
Scores refreshed
November 20th, 2026
13
Entries on its own leaderboard, 11 models we track

Qwen2.5-Coder-32B-Instruct leads with 91.7% among models released by January 2025 Qwen2.5-Coder-32B-Instruct leads with 91.7% among models released by January 2025

Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.

The newest model with a score here came out in January 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 Qwen2.5-Coder-32B-Instruct q4_0 91.7%
  2. 2 Qwen2.5-72B HF 85.4%
  3. 3 Llama 3.1-70B HF 80.9%
  4. 4 Llama 3.3 70B Instruct HF 78.9%
  5. 5 Phi-4 Q4_0 74.9%
  6. 6 Codestral Q6_K 68%
  7. 7 Viper-Coder-v1.7-Vsm6.Q8_0.gguf Q8_0 64.6%
  8. 8 FluentlyLM-Prinum.Q4_K_M.gguf Q4_K_M 40%
  9. 9 Llama 3.1-8B HF 34.9%
  10. 10 DeepSeek Coder 33B Q4_0 28.6%
  11. 11 DeepSeek-Coder-V2-Lite-Instruct Q6_K 26.3%
  12. 12 Qwen2.5-32B Q4_K_L 20.3%
  13. 13 Llama 3-8B Q4_0 19.7%

Source: Official AI Engineering Lab model creation leaderboard

#EntryScoreSettingDateSource
1 Qwen2.5-Coder-32B-Instruct 91.7% q4_0 Nov 2024 Official leaderboard
2 Qwen2.5-72B 85.4% HF Sep 2024 Official leaderboard
3 Llama 3.1-70B 80.9% HF Jul 2024 Official leaderboard
4 Llama 3.3 70B Instruct 78.9% HF Dec 2024 Official leaderboard
5 Phi-4 74.9% Q4_0 Dec 2024 Official leaderboard
6 Codestral 68% Q6_K Jan 2025 Official leaderboard
7 Viper-Coder-v1.7-Vsm6.Q8_0.gguf 64.6% Q8_0 Official leaderboard
8 FluentlyLM-Prinum.Q4_K_M.gguf 40% Q4_K_M Official leaderboard
9 Llama 3.1-8B 34.9% HF Jul 2024 Official leaderboard
10 DeepSeek Coder 33B 28.6% Q4_0 Nov 2023 Official leaderboard
11 DeepSeek-Coder-V2-Lite-Instruct 26.3% Q6_K Jun 2024 Official leaderboard
12 Qwen2.5-32B 20.3% Q4_K_L Sep 2024 Official leaderboard
13 Llama 3-8B 19.7% Q4_0 Apr 2024 Official leaderboard
#ModelScoreSettingDateSource
1 Qwen2.5-Coder-32B-Instruct 91.7% q4_0 Nov 2024 Official leaderboard
2 Qwen2.5-72B 85.4% HF Sep 2024 Official leaderboard
3 Llama 3.1-70B 80.9% HF Jul 2024 Official leaderboard
4 Llama 3.3 70B Instruct 78.9% HF Dec 2024 Official leaderboard
5 Phi-4 74.9% Q4_0 Dec 2024 Official leaderboard
6 Codestral 68% Q6_K Jan 2025 Official leaderboard
7 Llama 3.1-8B 34.9% HF Jul 2024 Official leaderboard
8 DeepSeek Coder 33B 28.6% Q4_0 Nov 2023 Official leaderboard
9 DeepSeek-Coder-V2-Lite-Instruct 26.3% Q6_K Jun 2024 Official leaderboard
10 Qwen2.5-32B 20.3% Q4_K_L Sep 2024 Official leaderboard
11 Llama 3-8B 19.7% Q4_0 Apr 2024 Official leaderboard

Rankings by part

We read 35 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as flyingMassPoint, freeFallMassPoint and singleMassOscillator. Llama 3.1-70B leads 18 of them, and 6 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

flyingMassPoint

11 models we track have a score here.

  1. 1 DeepSeek Coder 33B ↑7 100%
  2. 2 Llama 3-8B ↑9 100%
  3. 3 Llama 3.1-70B 100%
  4. 4 Llama 3.1-8B ↑3 100%
  5. 5 Qwen2.5-72B ↓3 100%
  6. 6 Qwen2.5-Coder-32B-Instruct ↓5 100%
  7. 7 Llama 3.3 70B Instruct ↓3 100%
  8. 8 Phi-4 ↓3 100%
  9. 9 Codestral ↓3 100%
  10. 10 Qwen2.5-32B 90%

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

Qwen
Own leaderboard Passed 85% in Apr 2025, by Qwen2.5-Coder-32B-Instruct

Source: Official AI Engineering Lab model creation leaderboard

Qwen2.5-72B on the official leaderboard passed 85% in September 2024

Models released before it came out already scored that high. All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2024 2025 Near the ceiling (85%) Qwen2.5-Coder-32B-Instruct Qwen2.5-72B Llama 3.1-70B DeepSeek Coder 33B 0% 20% 40% 60% 80% 100% 2024 2025 Near the ceiling (85%) Qwen2.5-Coder-32B-Instruct Qwen2.5-72B

Source: Official AI Engineering Lab model creation 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.

Versions

More in Coding

All 523 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

All 8 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 2 pass 3 notes
  • Pass Independent runs Scores from the official leaderboard (13 entries).
  • Pass Models scored 11 models we track, plus 2 other entries on its leaderboard.
  • Note Newest score The newest came 22 months ago, in January 2025, for Codestral. That’s expected for a test replaced by MecEng.
  • Note Frontier models None of the top 10 has a score, which is fine for a retired test. The strongest tested is Phi-4 (#160).
  • Note Headroom Saturated. The best is 91.7%, past 85% since September 2024.
Leaderboard import 3 pass
  • Pass Importer We read the board on github.com: 12 rows for models we track and 2 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.

How we check