Benchmark Almanac

SquadStack Conversational LLM Benchmark for Voice Agents

Maya V1 96.7% tops its leaderboard Gemma 4 31B IT 92.5% leads the models we track
Not in model cards

Models write the next reply or tool call from the history of an Indian outbound phone call. A language model grades each response against seven criteria.

SquadStack ranks 11 systems on histories from 471 calls. It evaluated 9,962 turn and prompt pairs. Some systems lack a verdict on some turns.

Released
November 2026
Kind
Customer service
Level
University
Scored
Graded by a model
Answers
Open-ended
Tasks
Kept private
Measures
Capability
In and out
Text → Text, Actions
Interaction
Single turn
Items from
Real work and use
Refresh
One fixed set
Run by
A host
Strongest model tested
DeepSeek V4.1 Flash #34 on the capability index
Last model added
September 2026
Scores refreshed
November 20th, 2026
11
Entries on its own leaderboard, 9 models we track

Maya V1 tops the leaderboard with 96.7% Gemma 4 31B IT leads with 92.5% among models released by September 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. The best model we track is Gemma 4 31B IT, at 92.5%. This view shows only models we track, each at its best.

The newest model with a score here came out in September 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 Maya V1 96.7%
  2. 2 Gemma 4 31B IT 92.5%
  3. 3 DeepSeek V4.1 Flash 92.2%
  4. 4 GPT-6 Luna 89%
  5. 5 Qwen 3.8 27B 83.8%
  6. 6 Gemma 4 26B A4B 81.7%
  7. 7 GPT-4.1 77.5%
  8. 8 Gnani Evon V3.3 30B A3B 76%
  9. 9 Sarvam 105B 72.4%
  10. 10 Nemotron 3.5 Lightning 30B-A3B 71.3%
  11. 11 Sarvam 30B 65.7%

Source: Official SquadStack Conversational LLM Benchmark for Voice Agents leaderboard

#EntryScore ±SettingDateSource
1 Maya V1 96.7% 96.4 to 97 Official leaderboard
2 Gemma 4 31B IT 92.5% ±0.4 Apr 2026 Official leaderboard
3 DeepSeek V4.1 Flash 92.2% 91.8 to 92.5 Sep 2026 Official leaderboard
4 GPT-6 Luna 89% 88.5 to 89.5 Sep 2026 Official leaderboard
5 Qwen 3.8 27B 83.8% 83.1 to 84.4 Aug 2026 Official leaderboard
6 Gemma 4 26B A4B 81.7% ±0.9 Apr 2026 Official leaderboard
7 GPT-4.1 77.5% 76.4 to 78.5 Apr 2025 Official leaderboard
8 Gnani Evon V3.3 30B A3B open weights 76% ±0.9 Official leaderboard
9 Sarvam 105B 72.4% ±0.9 Mar 2026 Official leaderboard
10 Nemotron 3.5 Lightning 30B-A3B 71.3% ±1 Aug 2026 Official leaderboard
11 Sarvam 30B 65.7% ±1 Mar 2026 Official leaderboard
#ModelScore ±SettingDateSource
1 Gemma 4 31B IT 92.5% ±0.4 Apr 2026 Official leaderboard
2 DeepSeek V4.1 Flash 92.2% 91.8 to 92.5 Sep 2026 Official leaderboard
3 GPT-6 Luna 89% 88.5 to 89.5 Sep 2026 Official leaderboard
4 Qwen 3.8 27B 83.8% 83.1 to 84.4 Aug 2026 Official leaderboard
5 Gemma 4 26B A4B 81.7% ±0.9 Apr 2026 Official leaderboard
6 GPT-4.1 77.5% 76.4 to 78.5 Apr 2025 Official leaderboard
7 Sarvam 105B 72.4% ±0.9 Mar 2026 Official leaderboard
8 Nemotron 3.5 Lightning 30B-A3B 71.3% ±1 Aug 2026 Official leaderboard
9 Sarvam 30B 65.7% ±1 Mar 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 Passed 85% in Nov 2026, by Gemma 4 31B IT

Source: Official SquadStack Conversational LLM Benchmark for Voice Agents leaderboard

Gemma 4 31B IT on the official leaderboard passed 85% in April 2026

Models released before it came out already scored that high. 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%) Gemma 4 31B IT GPT-4.1 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Gemma 4 31B IT GPT-4.1

Source: Official SquadStack Conversational LLM Benchmark for Voice Agents 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 The tasks are private.
  • Pass Confidence Checked against a primary source.
Scores 3 pass 2 notes
  • Pass Independent runs Scores from the official leaderboard (11 entries).
  • Pass Models scored 9 models we track, plus 2 other entries on its leaderboard.
  • Pass Newest score The newest is from September 2026, for GPT-6 Luna.
  • Note Frontier models None of the top 10 has a score, which is fine for a retired test. The strongest tested is DeepSeek V4.1 Flash (#34).
  • Note Headroom Saturated. The best is 92.5%, past 85% since April 2026.
Leaderboard import 3 pass
  • Pass Importer We read the board on huggingface.co: 9 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