Benchmark Almanac

PDEAgent-Bench

Gemini 3.1 Pro 54.1% leads since May 2026
Not in model cards

Models write finite element solvers for partial differential equations. A solver must run successfully, meet a numerical error tolerance, and finish within its time limit.

The authors rank five models and three agents across three solver libraries. The MUSE team separately evaluated its agent on 451 held-out cases.

Released
May 2026
Kind
Scientific code
Level
Expert
Scored
Checked automatically
Answers
Code
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Code, Actions
Interaction
Agentic
Environment
Code sandbox
Refresh
One fixed set
Run by
Anyone
In views
Agents
Strongest model tested
GPT-5.4 #28 on the capability index
Last model added
April 2026
Scores refreshed
November 20th, 2026
7
Entries on its own leaderboard, 5 models we track

Gemini 3.1 Pro leads with 54.1% among models released by April 2026 Gemini 3.1 Pro leads with 54.1% among models released by April 2026

Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran on its own with several settings, we keep its best. Every run is in the next chart. Agents other companies built around a model get a pale shade of that model's lab color. Gray marks blends and entries not in our model list, such as fine-tunes and older models. None of these count toward leaders. Agents built around a model, such as CodePDE + GPT-5.4, are their own entries. We don't count them as the model's score. This view shows only models we track, each at its best.

The newest model with a score here came out in April 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 Gemini 3.1 Pro DOLFINx 54.1%
  2. 2 Claude Opus 4.7 DOLFINx 47.8%
  3. 3 GPT-5.4 DOLFINx 46%
  4. 4 CodePDE + GPT-5.4 DOLFINx 44.7%
  5. 5 OpenHands + GPT-5.4 DOLFINx 43.6%
  6. 6 Qwen 3.6 Plus DOLFINx 23.4%
  7. 7 DeepSeek-V3.2 DOLFINx 4.20%

Source: Official PDEAgent-Bench leaderboard

#EntryScoreSettingDateSource
1 Gemini 3.1 Pro 54.1% DOLFINx Feb 2026 Official leaderboard
2 Claude Opus 4.7 47.8% DOLFINx Apr 2026 Official leaderboard
3 GPT-5.4 46% DOLFINx Mar 2026 Official leaderboard
4 CodePDE + GPT-5.4 44.7% DOLFINx Mar 2026 Official leaderboard
5 OpenHands + GPT-5.4 43.6% DOLFINx Mar 2026 Official leaderboard
6 Qwen 3.6 Plus 23.4% DOLFINx Mar 2026 Official leaderboard
7 DeepSeek-V3.2 4.20% DOLFINx Dec 2025 Official leaderboard
#ModelScoreSettingDateSource
1 Gemini 3.1 Pro 54.1% DOLFINx Feb 2026 Official leaderboard
2 Claude Opus 4.7 47.8% DOLFINx Apr 2026 Official leaderboard
3 GPT-5.4 46% DOLFINx Mar 2026 Official leaderboard
4 Qwen 3.6 Plus 23.4% DOLFINx Mar 2026 Official leaderboard
5 DeepSeek-V3.2 4.20% DOLFINx Dec 2025 Official leaderboard

Rankings by part

We read 11 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Poisson, Helmholtz and Biharmonic. Gemini 3.1 Pro leads 7 of them, and Claude Opus 4.7 leads the other 4. Arrows show how far a model moved from its overall rank among the models we track.

Poisson

5 models we track have a score here.

  1. 1 Claude Opus 4.7 ↑1 54.9%
  2. 2 Gemini 3.1 Pro ↓1 51.6%
  3. 3 GPT-5.4 47.3%
  4. 4 Qwen 3.6 Plus 27.5%
  5. 5 DeepSeek-V3.2 14.3%

The best score, 54.1%, costs $0.04 a task

Runs on the cost frontier score higher than any run that's cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

DeepSeek Google DeepMind Qwen Cost frontier One model at each effort level
0% 20% 40% 60% $0.001 $0.01 $0.10 $1.00 Cost per task, log scale Gemini 3.1 Pro, DOLFINx Qwen 3.6 Plus, DOLFINx DeepSeek-V3.2, DOLFINx 0% 20% 40% 60% $0.001 $0.01 $0.10 $1.00 Cost per task, log scale Gemini 3.1 Pro, DOLFINx Qwen 3.6 Plus, DOLFINx

Source: Official PDEAgent-Bench leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Gemini 3.1 Pro DOLFINx 54.1% $0.04
Claude Opus 4.7 DOLFINx 47.8% $0.09
GPT-5.4 DOLFINx 46% $0.06
Claude Opus 4.7 Firedrake 45.9% $0.09
CodePDE + GPT-5.4 DOLFINx CodePDE 44.7% $0.06
OpenHands + GPT-5.4 DOLFINx OpenHands 43.6% $0.15
GPT-5.4 Firedrake 38% $0.06
Gemini 3.1 Pro Firedrake 32.4% $0.04
Gemini 3.1 Pro deal.II 32.3% $0.04
OpenHands + GPT-5.4 Firedrake OpenHands 28.8% $0.15
Claude Opus 4.7 deal.II 28.1% $0.09
CodePDE + GPT-5.4 Firedrake CodePDE 25.3% $0.06
CodePDE + GPT-5.4 deal.II CodePDE 24.6% $0.06
Qwen 3.6 Plus DOLFINx 23.4% $0.02
Qwen 3.6 Plus Firedrake 22.9% $0.02
GPT-5.4 deal.II 19.2% $0.06
Qwen 3.6 Plus deal.II 16.9% $0.02
OpenHands + GPT-5.4 deal.II OpenHands 16.6% $0.15
GPT-5.4 Firedrake Mini-SWE-Agent 12.9% $0.07
DeepSeek-V3.2 DOLFINx 4.20% $0.003
GPT-5.4 DOLFINx Mini-SWE-Agent 3.70% $0.07
DeepSeek-V3.2 Firedrake 2.70% $0.003
DeepSeek-V3.2 deal.II 0% $0.003
GPT-5.4 deal.II Mini-SWE-Agent 0% $0.07

The best score, 54.1%, costs $0.04 a task

Runs on the cost frontier score higher than any run that's cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

DeepSeek Google DeepMind Qwen Cost frontier One model at each effort level
0% 20% 40% 60% $0.001 $0.002 $0.005 $0.01 $0.02 $0.05 $0.10 Cost per task, log scale Gemini 3.1 Pro, DOLFINx Qwen 3.6 Plus, DOLFINx DeepSeek-V3.2, DOLFINx 0% 20% 40% 60% $0.001 $0.01 $0.10 Cost per task, log scale Gemini 3.1 Pro, DOLFINx Qwen 3.6 Plus, DOLFINx

Source: Official PDEAgent-Bench leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Gemini 3.1 Pro DOLFINx 54.1% $0.04
Claude Opus 4.7 DOLFINx 47.8% $0.09
GPT-5.4 DOLFINx 46% $0.06
Claude Opus 4.7 Firedrake 45.9% $0.09
GPT-5.4 Firedrake 38% $0.06
Gemini 3.1 Pro Firedrake 32.4% $0.04
Gemini 3.1 Pro deal.II 32.3% $0.04
Claude Opus 4.7 deal.II 28.1% $0.09
Qwen 3.6 Plus DOLFINx 23.4% $0.02
Qwen 3.6 Plus Firedrake 22.9% $0.02
GPT-5.4 deal.II 19.2% $0.06
Qwen 3.6 Plus deal.II 16.9% $0.02
GPT-5.4 Firedrake Mini-SWE-Agent 12.9% $0.07
DeepSeek-V3.2 DOLFINx 4.20% $0.003
GPT-5.4 DOLFINx Mini-SWE-Agent 3.70% $0.07
DeepSeek-V3.2 Firedrake 2.70% $0.003
DeepSeek-V3.2 deal.II 0% $0.003
GPT-5.4 deal.II Mini-SWE-Agent 0% $0.07

Number one over time

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

Google DeepMind
Own leaderboard Not tested since May 2026

Source: Official PDEAgent-Bench leaderboard

Scores over time

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% 2026 Near the ceiling (85%) Gemini 3.1 Pro DeepSeek-V3.2 0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Gemini 3.1 Pro DeepSeek-V3.2

Source: Official PDEAgent-Bench 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

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

9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 7 months ago, in April 2026, for Claude Opus 4.7. 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 official leaderboard (7 entries).
  • Pass Models scored 5 models we track, plus 2 other entries on its leaderboard.
  • Warning Newest score The newest came 7 months ago, in April 2026, for Claude Opus 4.7.
  • Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.4 (#28).
  • Pass Headroom The best is 54.1%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on zeroeclipse00.github.io: 24 rows for models we track 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