Benchmark Almanac

Taste-Bench

GPT-5.6 Sol 59.7% leads since September 2026
Not in model cards

Models choose the better next step at 502 decision points in software engineering and research trajectories. A question scores only when both option orders receive the correct answer. The overall score equally weights engineering and research.

Dataset access requires approval to limit training contamination. Later trajectory outcomes supply labels. Request errors and unparseable responses count as wrong. The paired-order protocol gives random guessing an expected score of 25 percent.

Released
September 2026
Kind
Machine learning engineering
Level
Expert
Scored
Checked automatically
Answers
Multiple choice
Tasks
Kept private
Measures
Capability
Language
English
In and out
Text, Code → Text
Interaction
Single turn
Items from
Real work and use
Refresh
One fixed set
Run by
Anyone
Strongest model tested
Claude Opus 5 #11 on the capability index
Last model added
July 2026
Scores refreshed
November 18th, 2026

In the authors’ words

LLM agents increasingly work on long-horizon tasks, and the decisions they make along the way, such as which hypothesis to test or which implementation to build on, determine the outcome of the whole run. Making these decisions well is becoming a key capability for both engineering and research agents. We refer to the ability to make good long-horizon decisions as the taste of an agent. While existing benchmarks measure the end-to-end success of agents on long-horizon tasks, none of them measures the taste of an agent. To address this problem, we build Taste-Bench, a benchmark of taste questions constructed automatically from trajectories that agents produced in engineering and research tasks. Each question presents a decision fork, a point in a trajectory where multiple directions are available and one of them leads to a better outcome, and the evaluated model chooses among these directions without seeing what happens after the fork. We mine these forks automatically from parallel attempts at the same task and from detours inside a single trajectory, without needing human annotation. We evaluate frontier models on Taste-Bench and find that the best model answers only 59.7% of the questions correctly. We further find that forks whose deciding evidence appears later in the trajectory are much harder for every model, and that a larger reasoning budget does not improve the accuracy. Finally, we show that taste can be trained. We distill the judgment of a teacher that has seen the outcome into a student model, and the student makes better decisions on unseen tasks and improves end-to-end success on held-out SWE-bench Pro tasks.

Abstract of The Tasteful Agent: Measuring and Improving Taste in Long-Horizon Tasks (arXiv:2609.25804, CC0)
14
Models scored on its own leaderboard

GPT-5.6 Sol leads with 59.7% among models released by July 2026

Best score per model from the benchmark's own leaderboard, read on November 18th, 2026. Where a model ran with several settings or harnesses, we keep its best.

The newest model with a score here came out in July 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 GPT-5.6 Sol 59.7%
  2. 2 GPT-5.5 59.5%
  3. 3 Claude Opus 5 55.5%
  4. 4 Grok 4.5 54.6%
  5. 5 GPT-5.6 Terra 54%
  6. 6 GLM-5.2 53.9%
  7. 7 Claude Sonnet 5 51.6%
  8. 8 GPT-5.6 Luna 49%
  9. 9 MiniMax-M3 45.3%
  10. 10 DeepSeek-V4-Flash 43.3%
  11. 11 GPT-5.4 Mini 40.1%
  12. 12 Mistral Medium 3.5 37.7%
  13. 13 GPT-5.4 Nano 36.6%
  14. 14 Grok 4.20 reasoning 15.7%

Source: Official Taste-Bench leaderboard

#ModelScoreSettingDateSource
1 GPT-5.6 Sol 59.7% Jul 2026 Official leaderboard
2 GPT-5.5 59.5% Apr 2026 Official leaderboard
3 Claude Opus 5 55.5% Jul 2026 Official leaderboard
4 Grok 4.5 54.6% Jul 2026 Official leaderboard
5 GPT-5.6 Terra 54% Jul 2026 Official leaderboard
6 GLM-5.2 53.9% Jun 2026 Official leaderboard
7 Claude Sonnet 5 51.6% Jun 2026 Official leaderboard
8 GPT-5.6 Luna 49% Jul 2026 Official leaderboard
9 MiniMax-M3 45.3% Jun 2026 Official leaderboard
10 DeepSeek-V4-Flash 43.3% Apr 2026 Official leaderboard
11 GPT-5.4 Mini 40.1% Mar 2026 Official leaderboard
12 Mistral Medium 3.5 37.7% Apr 2026 Official leaderboard
13 GPT-5.4 Nano 36.6% Mar 2026 Official leaderboard
14 Grok 4.20 15.7% reasoning Feb 2026 Official leaderboard

Rankings by part

We read 4 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Detour engineering, Detour research and Parallel engineering. GPT-5.6 Sol leads 2 of them, and 2 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

Detour engineering

14 models we track have a score here.

  1. 1 GPT-5.6 Sol 48.1%
  2. 2 Grok 4.5 ↑2 47.7%
  3. 3 GPT-5.5 ↓1 47.4%
  4. 4 Mistral Medium 3.5 ↑8 40.6%
  5. 5 GPT-5.6 Terra 40.6%
  6. 6 GLM-5.2 40.2%
  7. 7 Claude Sonnet 5 36.1%
  8. 8 Claude Opus 5 ↓5 35.3%
  9. 9 MiniMax-M3 34.6%
  10. 10 GPT-5.6 Luna ↓2 32.3%

Number one over time

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

OpenAI
Own leaderboard Not tested since Sep 2026
GPT-5.6 Sol 59.7%

Source: Official Taste-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%) GPT-5.6 Sol GPT-5.5 GPT-5.4 Mini Grok 4.20 0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) GPT-5.6 Sol GPT-5.4 Mini

Source: Official Taste-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 AI research

All 164 in AI research
  • PaperBench In use Agent replicates ICML 2024 papers from scratch. LLM judge grades against author-written hierarchical rubrics. Qwen 3.8 Max 93%
  • MLS-Bench In use Agent must invent one algorithmic change in an ML research scaffold whose gain transfers across settings. Scored relative to human baselines. Claude Fable 5.1 50.3%
  • PostTrainBench 1.2 In use CLI agent gets a base model, one H100 and a time budget to post-train it for a target benchmark. Scored on normalized gain. Claude Fable 5.1 44.6%
  • MLE-bench In use Kaggle-style ML engineering competitions. Agent trains models and is scored against competition leaderboards. GPT-6.2 Astra 96.1%
  • PMPP-Hard In use Agents write GPU kernels for 69 tasks that must be correct, and on 48 of them also run within 1.25 times the reference time. The score is the area under the solved-tasks curve over a 10-hour budget, out of 690 task-hours. Kimi K3 0.71
  • KernelBench In use Replace PyTorch operators with faster custom GPU kernels. Scored on correctness and speedup. MiniMax-M3 28.8%

Data audit

All 10 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access The tasks are private.
  • Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
  • Pass Independent runs Scores from the official leaderboard (14 entries).
  • Pass Models scored 14 models we track.
  • Pass Newest score The newest is from July 2026, for Claude Opus 5.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Claude Opus 5 (#11).
  • Pass Headroom The best is 59.7%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on github.com: 14 rows for models we track at the last read.
  • Pass Last read November 18th, 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