Benchmark Almanac

KRYT Benchmark

Gemini 2.5 Pro Jun 2025 32% leads since October 2026
Not in model cards

Models decide refund requests from transaction limits and event records. The grader checks policy compliance, valid actions, and consistency across rounds.

Kaggle publishes scores for 43 distinct model versions.

Made by
vaibhav lalwani
Released
October 2026
Kind
Customer service
Level
Everyday
Scored
Checked automatically
Answers
Short answer
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Text
Interaction
Single turn
Refresh
One fixed set
Run by
Anyone
Strongest model tested
GPT-6 Astra #4 on the capability index
Last model added
October 2026
Scores refreshed
November 20th, 2026
42
Entries on its own leaderboard, 40 models we track

23 models share the top score of 32% 23 models share the top score of 32%

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.

  1. 1 Gemini 2.5 Pro Jun 2025 32%
  2. 2 gpt-oss-20b 32%
  3. 3 Gemini 2.5 Flash Sep 2025 32%
  4. 4 Claude Haiku 4.5 32%
  5. 5 Gemini 3 Flash 32%
  6. 6 Claude Opus 4.6 32%
  7. 7 GLM-5 32%
  8. 8 Claude Sonnet 4.6 32%
  9. 9 Grok 4.20 reasoning 32%
  10. 10 Gemini 3.1 Pro 32%
  11. 11 Gemini 3.1 Flash-Lite 32%
  12. 12 GPT-5.4 32%
  13. 13 Gemma 4 26B A4B 32%
  14. 14 Gemma 4 31B IT 32%
  15. 15 Gemini 3.5 Flash 32%
  16. 16 GPT-5.6 Luna 32%
  17. 17 GPT-5.6 Sol 32%
  18. 18 Gemini 3.6 Flash 32%
  19. 19 Gemini 3.8 Flash 32%
  20. 20 GPT-6 Luna 32%
  21. 21 Claude Sonnet 5.5 32%
  22. 22 GPT-6.1 Sol 32%
  23. 23 Claude Haiku 5.5 32%
  24. 24 gpt-oss-120b 31%
  25. 25 Qwen 3 235B A22B Instruct 2506 31%
  26. 26 DeepSeek-R1 May 2025 28%
  27. 27 GPT-5.4 Nano 28%
  28. 28 Qwen3-Coder-480B-A35B 27%
  29. 29 GPT-5.4 Mini 27%
  30. 30 Gemini 3.5 Flash-Lite 27%
  31. 31 Qwen 3 Next 80B Instruct 27%
  32. 32 Claude Sonnet 5 26%
  33. 33 Claude Opus 4.7 25%
  34. 34 GPT-5.5 25%
  35. 35 Claude Opus 5.5 25%
  36. 36 Claude Opus 4.8 24%
  37. 37 GPT-5.6 Terra 24%
  38. 38 Claude Opus 5 24%
  39. 39 GPT-6 Sol 24%
  40. 40 Grok 4.5 23%
  41. 41 Grok 4.6 23%
  42. 42 GPT-6 Astra 23%

Source: Official KRYT Benchmark leaderboard

Source: Official KRYT Benchmark leaderboard

#EntryScoreSettingReleasedRun onSource
1 Gemini 2.5 Pro Jun 2025 32% Jun 2025 Oct 2026 Official leaderboard
2 gpt-oss-20b 32% Aug 2025 Oct 2026 Official leaderboard
3 Gemini 2.5 Flash Sep 2025 32% Sep 2025 Oct 2026 Official leaderboard
4 Claude Haiku 4.5 32% Oct 2025 Oct 2026 Official leaderboard
5 Gemini 3 Flash 32% Dec 2025 Oct 2026 Official leaderboard
6 Claude Opus 4.6 32% Feb 2026 Oct 2026 Official leaderboard
7 GLM-5 32% Feb 2026 Oct 2026 Official leaderboard
8 Claude Sonnet 4.6 32% Feb 2026 Oct 2026 Official leaderboard
9 Grok 4.20 32% reasoning Feb 2026 Oct 2026 Official leaderboard
10 Gemini 3.1 Pro 32% Feb 2026 Oct 2026 Official leaderboard
11 Gemini 3.1 Flash-Lite 32% Mar 2026 Oct 2026 Official leaderboard
12 GPT-5.4 32% Mar 2026 Oct 2026 Official leaderboard
13 Gemma 4 26B A4B 32% Apr 2026 Oct 2026 Official leaderboard
14 Gemma 4 31B IT 32% Apr 2026 Oct 2026 Official leaderboard
15 Gemini 3.5 Flash 32% May 2026 Oct 2026 Official leaderboard
16 GPT-5.6 Luna 32% Jul 2026 Oct 2026 Official leaderboard
17 GPT-5.6 Sol 32% Jul 2026 Oct 2026 Official leaderboard
18 Gemini 3.6 Flash 32% Jul 2026 Oct 2026 Official leaderboard
19 Gemini 3.8 Flash 32% Sep 2026 Oct 2026 Official leaderboard
20 GPT-6 Luna 32% Sep 2026 Oct 2026 Official leaderboard
21 Claude Sonnet 5.5 32% Sep 2026 Oct 2026 Official leaderboard
22 GPT-6.1 Sol 32% Sep 2026 Oct 2026 Official leaderboard
23 Claude Haiku 5.5 32% Oct 2026 Oct 2026 Official leaderboard
24 gpt-oss-120b 31% Aug 2025 Oct 2026 Official leaderboard
25 Qwen 3 235B A22B Instruct 2506 31% Oct 2026 Official leaderboard
26 DeepSeek-R1 May 2025 28% May 2025 Oct 2026 Official leaderboard
27 GPT-5.4 Nano 28% Mar 2026 Oct 2026 Official leaderboard
28 Qwen3-Coder-480B-A35B 27% Jul 2025 Oct 2026 Official leaderboard
29 GPT-5.4 Mini 27% Mar 2026 Oct 2026 Official leaderboard
30 Gemini 3.5 Flash-Lite 27% Jul 2026 Oct 2026 Official leaderboard
31 Qwen 3 Next 80B Instruct 27% Oct 2026 Official leaderboard
32 Claude Sonnet 5 26% Jun 2026 Oct 2026 Official leaderboard
33 Claude Opus 4.7 25% Apr 2026 Oct 2026 Official leaderboard
34 GPT-5.5 25% Apr 2026 Oct 2026 Official leaderboard
35 Claude Opus 5.5 25% Sep 2026 Oct 2026 Official leaderboard
36 Claude Opus 4.8 24% May 2026 Oct 2026 Official leaderboard
37 GPT-5.6 Terra 24% Jul 2026 Oct 2026 Official leaderboard
38 Claude Opus 5 24% Jul 2026 Oct 2026 Official leaderboard
39 GPT-6 Sol 24% Sep 2026 Oct 2026 Official leaderboard
40 Grok 4.5 23% Jul 2026 Oct 2026 Official leaderboard
41 Grok 4.6 23% Aug 2026 Oct 2026 Official leaderboard
42 GPT-6 Astra 23% Sep 2026 Oct 2026 Official leaderboard
#ModelScoreSettingReleasedRun onSource
1 Gemini 2.5 Pro Jun 2025 32% Jun 2025 Oct 2026 Official leaderboard
2 gpt-oss-20b 32% Aug 2025 Oct 2026 Official leaderboard
3 Gemini 2.5 Flash Sep 2025 32% Sep 2025 Oct 2026 Official leaderboard
4 Claude Haiku 4.5 32% Oct 2025 Oct 2026 Official leaderboard
5 Gemini 3 Flash 32% Dec 2025 Oct 2026 Official leaderboard
6 Claude Opus 4.6 32% Feb 2026 Oct 2026 Official leaderboard
7 GLM-5 32% Feb 2026 Oct 2026 Official leaderboard
8 Claude Sonnet 4.6 32% Feb 2026 Oct 2026 Official leaderboard
9 Grok 4.20 32% reasoning Feb 2026 Oct 2026 Official leaderboard
10 Gemini 3.1 Pro 32% Feb 2026 Oct 2026 Official leaderboard
11 Gemini 3.1 Flash-Lite 32% Mar 2026 Oct 2026 Official leaderboard
12 GPT-5.4 32% Mar 2026 Oct 2026 Official leaderboard
13 Gemma 4 26B A4B 32% Apr 2026 Oct 2026 Official leaderboard
14 Gemma 4 31B IT 32% Apr 2026 Oct 2026 Official leaderboard
15 Gemini 3.5 Flash 32% May 2026 Oct 2026 Official leaderboard
16 GPT-5.6 Luna 32% Jul 2026 Oct 2026 Official leaderboard
17 GPT-5.6 Sol 32% Jul 2026 Oct 2026 Official leaderboard
18 Gemini 3.6 Flash 32% Jul 2026 Oct 2026 Official leaderboard
19 Gemini 3.8 Flash 32% Sep 2026 Oct 2026 Official leaderboard
20 GPT-6 Luna 32% Sep 2026 Oct 2026 Official leaderboard
21 Claude Sonnet 5.5 32% Sep 2026 Oct 2026 Official leaderboard
22 GPT-6.1 Sol 32% Sep 2026 Oct 2026 Official leaderboard
23 Claude Haiku 5.5 32% Oct 2026 Oct 2026 Official leaderboard
24 gpt-oss-120b 31% Aug 2025 Oct 2026 Official leaderboard
25 DeepSeek-R1 May 2025 28% May 2025 Oct 2026 Official leaderboard
26 GPT-5.4 Nano 28% Mar 2026 Oct 2026 Official leaderboard
27 Qwen3-Coder-480B-A35B 27% Jul 2025 Oct 2026 Official leaderboard
28 GPT-5.4 Mini 27% Mar 2026 Oct 2026 Official leaderboard
29 Gemini 3.5 Flash-Lite 27% Jul 2026 Oct 2026 Official leaderboard
30 Claude Sonnet 5 26% Jun 2026 Oct 2026 Official leaderboard
31 Claude Opus 4.7 25% Apr 2026 Oct 2026 Official leaderboard
32 GPT-5.5 25% Apr 2026 Oct 2026 Official leaderboard
33 Claude Opus 5.5 25% Sep 2026 Oct 2026 Official leaderboard
34 Claude Opus 4.8 24% May 2026 Oct 2026 Official leaderboard
35 GPT-5.6 Terra 24% Jul 2026 Oct 2026 Official leaderboard
36 Claude Opus 5 24% Jul 2026 Oct 2026 Official leaderboard
37 GPT-6 Sol 24% Sep 2026 Oct 2026 Official leaderboard
38 Grok 4.5 23% Jul 2026 Oct 2026 Official leaderboard
39 Grok 4.6 23% Aug 2026 Oct 2026 Official leaderboard
40 GPT-6 Astra 23% Sep 2026 Oct 2026 Official leaderboard

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 1 leader

Source: Official KRYT Benchmark 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% 2025 2026 Near the ceiling (85%) Gemini 2.5 Pro 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Gemini 2.5 Pro

Source: Official KRYT Benchmark 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 Tool use

All 265 in Tool use
  • AutomationBench In use Agent completes end-to-end business workflows inside a simulated company exposing dozens of REST APIs. Graded on resulting state. Gemini 4 Neon 56.3%
  • MCP Atlas In use Multi-step workflows over real MCP servers. Scores tool-use competency on realistic tasks. Gemini 4 Argon 90.7%
  • BFCL V4 In use Models call functions and tools, including multi-turn agentic tasks with web search and memory. Checked against expected calls and outcomes. Qwen3.7-Max 75%
  • τ³-bench Banking In use Agents act as banking customer-service reps over many turns with a simulated user, following policy. Pass^1 on the final database state. Qwen 3.8 Max 55.1%
  • Toolathlon-Verified In use 108 long-horizon tasks across 32 apps through about 600 MCP tools. Execution-based checkers inspect final environment state. GLM-5.3-Flash 78.4%
  • DeepSearchQA In use Questions that need multi-step search and synthesis. Final answers graded against references. Kimi K3 95%

Data audit

All 11 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 5 pass
  • Pass Independent runs Scores from the official leaderboard (42 entries).
  • Pass Models scored 40 models we track, plus 2 other entries on its leaderboard.
  • Pass Newest score The newest is from October 2026, for Claude Haiku 5.5.
  • Pass Frontier models 4 of the top 10 have a score, led by GPT-6 Astra (#4).
  • Pass Headroom The best is 32%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on kaggle.com: 41 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