Benchmark Almanac

MetroLLM-Bench

Muse Glimmer 92.0 leads since August 2026
Not in model cards

Agents operate a transit kiosk using tools for routes, fares, and service disruptions. The first score checks their actions automatically. The composite also includes model judgments.

Continker publishes 25 ranked configurations on 238 held-out cases across six metro systems. Three rows use GPT-5.4 full at different reasoning settings.

Made by
Continker
Released
June 2026
Kind
Customer service
Level
Everyday
Scored
Checked automatically
Answers
Agent task
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Actions, Text
Interaction
Agentic
Items from
Real work and use
Refresh
One fixed set
Run by
Anyone
In views
Agents, Robotics
Strongest model tested
GPT-5.6 Sol #14 on the capability index
Last model added
August 2026
Scores refreshed
November 20th, 2026
23
Entries on its own leaderboard, 19 models we track

Muse Glimmer leads with 92.0 among models released by August 2026 Muse Glimmer leads with 92.0 among models released by August 2026

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.

The newest model with a score here came out in August 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 Muse Glimmer 92.0
  2. 2 Qwen 3.8 27B 91.8
  3. 3 Qwen3.6 27B 91.3
  4. 4 Qwen3.5-27B 90.6
  5. 5 GPT-5.6 Luna medium 90.6
  6. 6 GPT-5.4 xhigh 90.5
  7. 7 Qwen3.5-35B-A3B 89.9
  8. 8 GPT-5.6 Sol xhigh 89.8
  9. 9 Qwen3.5-27B + fine-tune 89.7
  10. 10 Qwen3.5-4B + fine-tune 2.6 GB, offline 89.1
  11. 11 Qwen3.5-9B + fine-tune 88.8
  12. 12 Qwen3.5-9B 88.0
  13. 13 Mistral Small 4 87.8
  14. 14 Qwen3.5-4B 87.3
  15. 15 glm-4.7-flash 86.7
  16. 16 GPT-5.4 Nano 86.6
  17. 17 GPT-5.4 Mini 86.4
  18. 18 Ministral 3 8B Instruct 85.7
  19. 19 Gemma 4 26B A4B 82.0
  20. 20 Qwen3.5-2B + fine-tune 79.7
  21. 21 Qwen3.5-2B 71.9
  22. 22 Qwen3.5-0.8B 60.0
  23. 23 Mistral NeMo 56.7

Source: Official MetroLLM-Bench leaderboard

#EntryScoreSettingDateSource
1 Muse Glimmer 92.0 Aug 2026 Official leaderboard
2 Qwen 3.8 27B 91.8 Aug 2026 Official leaderboard
3 Qwen3.6 27B 91.3 Apr 2026 Official leaderboard
4 Qwen3.5-27B 90.6 Feb 2026 Official leaderboard
5 GPT-5.6 Luna 90.6 medium Jul 2026 Official leaderboard
6 GPT-5.4 90.5 xhigh Mar 2026 Official leaderboard
7 Qwen3.5-35B-A3B 89.9 Feb 2026 Official leaderboard
8 GPT-5.6 Sol 89.8 xhigh Jul 2026 Official leaderboard
9 Qwen3.5-27B + fine-tune open weights 89.7 Official leaderboard
10 Qwen3.5-4B + fine-tune 2.6 GB, offline open weights 89.1 Official leaderboard
11 Qwen3.5-9B + fine-tune open weights 88.8 Official leaderboard
12 Qwen3.5-9B 88.0 Feb 2026 Official leaderboard
13 Mistral Small 4 87.8 Mar 2026 Official leaderboard
14 Qwen3.5-4B 87.3 Feb 2026 Official leaderboard
15 glm-4.7-flash 86.7 Jan 2026 Official leaderboard
16 GPT-5.4 Nano 86.6 Mar 2026 Official leaderboard
17 GPT-5.4 Mini 86.4 Mar 2026 Official leaderboard
18 Ministral 3 8B Instruct 85.7 Dec 2025 Official leaderboard
19 Gemma 4 26B A4B 82.0 Apr 2026 Official leaderboard
20 Qwen3.5-2B + fine-tune open weights 79.7 Official leaderboard
21 Qwen3.5-2B 71.9 Feb 2026 Official leaderboard
22 Qwen3.5-0.8B 60.0 Feb 2026 Official leaderboard
23 Mistral NeMo 56.7 Jul 2024 Official leaderboard
#ModelScoreSettingDateSource
1 Muse Glimmer 92.0 Aug 2026 Official leaderboard
2 Qwen 3.8 27B 91.8 Aug 2026 Official leaderboard
3 Qwen3.6 27B 91.3 Apr 2026 Official leaderboard
4 Qwen3.5-27B 90.6 Feb 2026 Official leaderboard
5 GPT-5.6 Luna 90.6 medium Jul 2026 Official leaderboard
6 GPT-5.4 90.5 xhigh Mar 2026 Official leaderboard
7 Qwen3.5-35B-A3B 89.9 Feb 2026 Official leaderboard
8 GPT-5.6 Sol 89.8 xhigh Jul 2026 Official leaderboard
9 Qwen3.5-9B 88.0 Feb 2026 Official leaderboard
10 Mistral Small 4 87.8 Mar 2026 Official leaderboard
11 Qwen3.5-4B 87.3 Feb 2026 Official leaderboard
12 glm-4.7-flash 86.7 Jan 2026 Official leaderboard
13 GPT-5.4 Nano 86.6 Mar 2026 Official leaderboard
14 GPT-5.4 Mini 86.4 Mar 2026 Official leaderboard
15 Ministral 3 8B Instruct 85.7 Dec 2025 Official leaderboard
16 Gemma 4 26B A4B 82.0 Apr 2026 Official leaderboard
17 Qwen3.5-2B 71.9 Feb 2026 Official leaderboard
18 Qwen3.5-0.8B 60.0 Feb 2026 Official leaderboard
19 Mistral NeMo 56.7 Jul 2024 Official leaderboard

2 models have held number one on the official leaderboard since its release in June 2026

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

Qwen Meta
Own leaderboard Not tested since Aug 2026

Source: Official MetroLLM-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
50.0 60.0 70.0 80.0 90.0 100.0 2025 2026 50.0 60.0 70.0 80.0 90.0 100.0 2025 2026

Source: Official MetroLLM-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 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 9 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 3 pass 2 notes
  • Pass Independent runs Scores from the official leaderboard (23 entries).
  • Pass Models scored 19 models we track, plus 4 other entries on its leaderboard.
  • Pass Newest score The newest is from August 2026, for Qwen 3.8 27B.
  • Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.6 Sol (#14).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on continker.ai: 21 rows for models we track and 4 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