Benchmark Almanac

OTelBench

Claude Opus 4.5 29% leads since January 2026
Not in model cards

Coding agents add OpenTelemetry tracing, metrics and logging to codebases in several languages. Tests check the instrumentation.

23 tasks and 14 models as of January 16th, 2026.

Made by
Quesma
Released
January 2026
Kind
Fixing real repositories
Level
Expert
Scored
Checked automatically
Answers
Code
Tasks
All public
Strongest model tested
GPT-5.2 #50 on the capability index
Last model added
December 2025
Scores refreshed
November 12th, 2026
14
Models scored on its own leaderboard

Claude Opus 4.5 leads with 29% among models released by December 2025

Best score per model from the benchmark's own leaderboard, read on November 12th, 2026. 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 December 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

14 models. Scroll sideways for the rest.

  1. 29.0 Claude Opus 4.5
  2. 26.1 GPT-5.2
  3. 21.7 Claude Sonnet 4.5
  4. 18.8 Gemini 3 Flash
  5. 15.9 Gemini 3 Pro
  6. 15.9 GPT-5.2 Codex
  7. 14.5 GPT-5.1
  8. 13.0 GLM-4.7
  9. 11.6 GPT-5.1-Codex-Max
  10. 11.6 DeepSeek-V3.2
  11. 7.25 Kimi K2 Thinking
  12. 5.80 Claude Haiku 4.5
  13. 4.35 Grok 4
  14. 2.90 Grok 4.1 Fast
  1. 1 Claude Opus 4.5 29%
  2. 2 GPT-5.2 26.1%
  3. 3 Claude Sonnet 4.5 21.7%
  4. 4 Gemini 3 Flash 18.8%
  5. 5 Gemini 3 Pro 15.9%
  6. 6 GPT-5.2 Codex 15.9%
  7. 7 GPT-5.1 14.5%
  8. 8 GLM-4.7 13%
  9. 9 GPT-5.1-Codex-Max 11.6%
  10. 10 DeepSeek-V3.2 11.6%
  11. 11 Kimi K2 Thinking 7.25%
  12. 12 Claude Haiku 4.5 5.80%
  13. 13 Grok 4 4.35%
  14. 14 Grok 4.1 Fast 2.90%

Source: Official OTelBench leaderboard

#ModelScoreSettingDateSource
1 Claude Opus 4.5 29% Nov 2025 Official leaderboard
2 GPT-5.2 26.1% Dec 2025 Official leaderboard
3 Claude Sonnet 4.5 21.7% Sep 2025 Official leaderboard
4 Gemini 3 Flash 18.8% Dec 2025 Official leaderboard
5 Gemini 3 Pro 15.9% Nov 2025 Official leaderboard
6 GPT-5.2 Codex 15.9% Dec 2025 Official leaderboard
7 GPT-5.1 14.5% Nov 2025 Official leaderboard
8 GLM-4.7 13% Dec 2025 Official leaderboard
9 GPT-5.1-Codex-Max 11.6% Nov 2025 Official leaderboard
10 DeepSeek-V3.2 11.6% Dec 2025 Official leaderboard
11 Kimi K2 Thinking 7.25% Nov 2025 Official leaderboard
12 Claude Haiku 4.5 5.80% Oct 2025 Official leaderboard
13 Grok 4 4.35% Jul 2025 Official leaderboard
14 Grok 4.1 Fast 2.90% Nov 2025 Official leaderboard

Speed vs quality

Pass rate against average generation time per task; the Pareto frontier marks models no other beats on both. The line joins the runs that no other run beats on both measures. We redrew this chart from the numbers on the official leaderboard.

0% 10% 20% 30% 05001,0001,500 Average time per task (s), lower is better Pass rate Grok 4: Pass rate 4.35%, average time per task 982 s Claude Haiku 4.5: Pass rate 5.8%, average time per task 511 s Kimi K2 Thinking: Pass rate 7.25%, average time per task 1,274 s GPT-5.1: Pass rate 14.5%, average time per task 920 s Gemini 3 Pro: Pass rate 15.9%, average time per task 485 s GPT-5.1-Codex-Max: Pass rate 11.6%, average time per task 1,095 s Grok 4.1 Fast: Pass rate 2.9%, average time per task 1,072 s DeepSeek-V3.2: Pass rate 11.6%, average time per task 1,312 s GPT-5.2: Pass rate 26.1%, average time per task 1,067 s GPT-5.2 Codex: Pass rate 15.9%, average time per task 1,300 s GLM-4.7: Pass rate 13%, average time per task 1,098 s Claude Sonnet 4.5: Pass rate 21.7%, average time per task 682 s Claude Opus 4.5: Pass rate 29%, average time per task 730 s Gemini 3 Flash: Pass rate 18.8%, average time per task 339 s Claude Opus 4.5Claude Sonnet 4.5Gemini 3 Flash 0% 10% 20% 30% 05001,0001,500 Average time per task (s), lower is better Pass rate Grok 4: Pass rate 4.35%, average time per task 982 s Claude Haiku 4.5: Pass rate 5.8%, average time per task 511 s Kimi K2 Thinking: Pass rate 7.25%, average time per task 1,274 s GPT-5.1: Pass rate 14.5%, average time per task 920 s Gemini 3 Pro: Pass rate 15.9%, average time per task 485 s GPT-5.1-Codex-Max: Pass rate 11.6%, average time per task 1,095 s Grok 4.1 Fast: Pass rate 2.9%, average time per task 1,072 s DeepSeek-V3.2: Pass rate 11.6%, average time per task 1,312 s GPT-5.2: Pass rate 26.1%, average time per task 1,067 s GPT-5.2 Codex: Pass rate 15.9%, average time per task 1,300 s GLM-4.7: Pass rate 13%, average time per task 1,098 s Claude Sonnet 4.5: Pass rate 21.7%, average time per task 682 s Claude Opus 4.5: Pass rate 29%, average time per task 730 s Gemini 3 Flash: Pass rate 18.8%, average time per task 339 s Claude Opus 4.5Claude Sonnet 4.5Gemini 3 Flash
ModelSettingPass rateAverage time per task
Claude Opus 4.5 29% 730.01 s
GPT-5.2 26.1% 1,067 s
Claude Sonnet 4.5 21.7% 682.34 s
Gemini 3 Flash 18.8% 338.65 s
Gemini 3 Pro 15.9% 485.34 s
GPT-5.2 Codex 15.9% 1,300 s
GPT-5.1 14.5% 919.64 s
GLM-4.7 13% 1,098 s
GPT-5.1-Codex-Max 11.6% 1,095 s
DeepSeek-V3.2 11.6% 1,312 s
Kimi K2 Thinking 7.2% 1,274 s
Claude Haiku 4.5 5.8% 510.75 s
Grok 4 4.3% 981.95 s
Grok 4.1 Fast 2.9% 1,072 s

The best score, 29%, costs $84 for the full run

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.

Anthropic OpenAI Google DeepMind Cost frontier One model at each effort level
0% 10% 20% 30% $1.00 $2.00 $5.00 $10 $20 $50 $100 Cost of the full run, log scale Claude Opus 4.5 GPT-5.2 Gemini 3 Flash 0% 10% 20% 30% $1.00 $10 $100 Cost of the full run, log scale Claude Opus 4.5 GPT-5.2

Source: Official OTelBench leaderboard

ModelEffortHarnessScoreCost of the full runOutput tokens
Claude Opus 4.5 29% $84
GPT-5.2 26.1% $48
Claude Sonnet 4.5 21.7% $69
Gemini 3 Flash 18.8% $7.41
Gemini 3 Pro 15.9% $33
GPT-5.2 Codex 15.9% $40
GPT-5.1 14.5% $30
GLM-4.7 13% $23
GPT-5.1-Codex-Max 11.6% $57
DeepSeek-V3.2 11.6% $12
Kimi K2 Thinking 7.25% $9.01
Claude Haiku 4.5 5.80% $30
Grok 4 4.35% $56
Grok 4.1 Fast 2.90% $9.85

Number one over time

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

Anthropic
Own leaderboard Not tested since Jan 2026
Claude Opus 4.5 29%

Source: Official OTelBench leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Claude Opus 4.5 Claude Sonnet 4.5 Grok 4 0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Claude Opus 4.5 Claude Sonnet 4.5

Source: Official OTelBench leaderboard

No lab model card reports it yet

We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.

More in Coding

All 282 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