AgentScript-Eval
Models write, debug, or optimize AgentScript definitions for Salesforce agents. A task passes when the program compiles and passes every execution test.
- Made by
- Salesforce
- Released
- November 2026
- Kind
- Function-level problems
- Level
- University
- Scored
- Checked automatically
- Answers
- Code
- Measures
- Capability
- Language
- English
- Interaction
- Agentic
- Environment
- Code sandbox
- Refresh
- One fixed set
- Run by
- A host
- In views
- Agents
- Strongest model tested
- GPT-5.5 #21 on the capability index
- Last model added
- June 2026
- Scores refreshed
- November 20th, 2026
Claude Opus 4.8 leads with 37% among models released by June 2026
Best score per model from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.
The newest model with a score here came out in June 2026, 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
Source: The AgentScript-Eval paper
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Opus 4.8 | 37% | Agentic, pass@1, Skills v3 (rewritten) | May 2026 | Official leaderboard |
| 2 | Claude Sonnet 5 | 28% | Agentic, pass@1, afcode (minimal CLI) | Jun 2026 | Official leaderboard |
| 3 | GPT-5.5 | 20% | Zero-shot direct, pass@3 | Apr 2026 | Official leaderboard |
| 4 | GPT-5 mini | 15% | Two-shot direct, pass@3 | Aug 2025 | Official leaderboard |
| 5 | | 15% | Zero-shot direct, pass@3 | Apr 2026 | Official leaderboard |
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 AgentScript-Eval paper
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| Claude Opus 4.8 | Agentic, pass@1 | Skills v3 (rewritten) | 37% | 440,000 | |
| Claude Opus 4.8 | Agentic, pass@1 | Baseline (no skills) | 33% | 835,000 | |
| Claude Opus 4.8 | Agentic, pass@1 | Skills v1 (inert docs) | 28% | 480,000 | |
| Claude Sonnet 5 | Agentic, pass@1 | afcode (minimal CLI) | 28% | 400,000 | |
| Claude Sonnet 5 | Agentic, pass@1 | Skills v1 (inert docs) | 27% | 1,125,000 | |
| Claude Opus 4.8 | Zero-shot direct, pass@3 | 25% | 9,900 | ||
| Claude Sonnet 5 | Agentic, pass@1 | Skills v3 (rewritten) | 23% | 914,000 | |
| Claude Opus 4.8 | Two-shot direct, pass@3 | 21% | 14,400 | ||
| Claude Opus 4.8 | Agentic, pass@1 | afcode (minimal CLI) | 21% | 293,000 | |
| GPT-5.5 | Zero-shot direct, pass@3 | 20% | 9,500 | ||
| GPT-5.5 | Two-shot direct, pass@3 | 20% | 11,000 | ||
| Claude Sonnet 5 | Agentic, pass@1 | Baseline (no skills) | 20% | 1,563,000 | |
| Claude Sonnet 5 | Two-shot direct, pass@3 | 16% | 9,800 | ||
| Claude Sonnet 5 | Agentic, pass@1 | Skills v2 (live plugin) | 16% | 1,123,000 | |
| GPT-5 mini | Two-shot direct, pass@3 | 15% | 10,700 | ||
| Zero-shot direct, pass@3 | 15% | 7,000 | |||
| GPT-5 mini | Zero-shot direct, pass@3 | 14% | 9,200 | ||
| Two-shot direct, pass@3 | 14% | 7,000 | |||
| Claude Opus 4.8 | Agentic, pass@1 | Skills v2 (live plugin) | 14% | 389,000 | |
| Claude Sonnet 5 | Zero-shot direct, pass@3 | 14% | 8,200 |
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: The AgentScript-Eval 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 AgentScript-Eval 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 Task access check. We don’t know whether the tasks are public. Every benchmark
Catalog record 1 warning 1 pass
- Warning Task access We don’t know whether the tasks are public.
- Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
- Pass Independent runs Scores from the benchmark’s paper (5 entries).
- Pass Models scored 5 models we track.
- Pass Newest score The newest is from June 2026, for Claude Sonnet 5.
- Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.5 (#21).
- Pass Headroom The best is 37%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on arxiv.org: 20 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.