SimulatorArena
Models act as students or writing clients. Researchers compare their messages and assistant ratings with those from people who used the same assistants.
The authors publish scores for 18 assistants and compare six simulator model releases. They collected 450 tutoring conversations and 459 writing conversations.
- Made by
- Georgia Institute of Technology and Microsoft
- Kind
- Multi-turn conversation
- Level
- School
- Scored
- Graded by a model
- Answers
- Agent task
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Multi-agent
- Items from
- Real work and use
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Agents
- Strongest model tested
- GPT-5 #71 on the capability index
- Last model added
- August 2025
- Scores refreshed
- November 20th, 2026
GPT-5 leads with 1.01 among models released by August 2025
Best score per model from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best.
The newest model with a score here came out in August 2025, 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
- 1.01 GPT-5
- 0.81 Claude 3.7 Sonnet
- 0.77 Claude Opus 4.1
- 0.51 GPT-4 Turbo
- 0.50 GPT-4o
- 0.49 Claude Sonnet 4
- 0.45 GPT-4.1
- 0.44
Phi-4
- 0.34 Claude 3.5 Sonnet
- 0.02 GPT-4o mini
- -0.03
Gemini 2.5 Flash
- -0.14
Gemini 2.5 Pro
- -0.16
Gemini 2.0 Flash
- -0.20
Mistral Large 2
- -0.29
Llama 3.3 70B
- -0.41
Llama 3.1-70B
- -1.56
Llama 3.1-8B
- -2.56
phi-3-medium 14B
- 1 GPT-5 Non-thinking 1.01
- 2 Claude 3.7 Sonnet Non-thinking 0.81
- 3 Claude Opus 4.1 Non-thinking 0.77
- 4 GPT-4 Turbo Apr 2024, Non-thinking 0.51
- 5 GPT-4o Nov 2024, Non-thinking 0.50
- 6 Claude Sonnet 4 Non-thinking 0.49
- 7 GPT-4.1 Non-thinking 0.45
- 8
Phi-4 Non-thinking 0.44
- 9 Claude 3.5 Sonnet Non-thinking 0.34
- 10 GPT-4o mini Non-thinking 0.02
- 11
Gemini 2.5 Flash Jun 2025, Non-thinking -0.03
- 12
Gemini 2.5 Pro Jun 2025, Non-thinking -0.14
- 13
Gemini 2.0 Flash Feb 2025, Non-thinking -0.16
- 14
Mistral Large 2 Jul 2024, Non-thinking -0.20
- 15
Llama 3.3 70B Non-thinking -0.29
- 16
Llama 3.1-70B Non-thinking -0.41
- 17
Llama 3.1-8B Non-thinking -1.56
- 18
phi-3-medium 14B Non-thinking -2.56
Source: Official SimulatorArena leaderboard
| # | Model | Score | Setting | Released | Run on | Source |
|---|---|---|---|---|---|---|
| 1 | GPT-5 | 1.01 | Non-thinking | Aug 2025 | Aug 2025 | Official leaderboard |
| 2 | Claude 3.7 Sonnet | 0.81 | Non-thinking | Feb 2025 | Aug 2025 | Official leaderboard |
| 3 | Claude Opus 4.1 | 0.77 | Non-thinking | Aug 2025 | Aug 2025 | Official leaderboard |
| 4 | GPT-4 Turbo Apr 2024 | 0.51 | Non-thinking | Apr 2024 | Aug 2025 | Official leaderboard |
| 5 | GPT-4o Nov 2024 | 0.50 | Non-thinking | Nov 2024 | Aug 2025 | Official leaderboard |
| 6 | Claude Sonnet 4 | 0.49 | Non-thinking | May 2025 | Aug 2025 | Official leaderboard |
| 7 | GPT-4.1 | 0.45 | Non-thinking | Apr 2025 | Aug 2025 | Official leaderboard |
| 8 | | 0.44 | Non-thinking | Dec 2024 | Aug 2025 | Official leaderboard |
| 9 | Claude 3.5 Sonnet | 0.34 | Non-thinking | Jun 2024 | Aug 2025 | Official leaderboard |
| 10 | GPT-4o mini | 0.02 | Non-thinking | Jul 2024 | Aug 2025 | Official leaderboard |
| 11 | | -0.03 | Non-thinking | Jun 2025 | Aug 2025 | Official leaderboard |
| 12 | | -0.14 | Non-thinking | Jun 2025 | Aug 2025 | Official leaderboard |
| 13 | | -0.16 | Non-thinking | Feb 2025 | Aug 2025 | Official leaderboard |
| 14 | | -0.20 | Non-thinking | Jul 2024 | Aug 2025 | Official leaderboard |
| 15 | | -0.29 | Non-thinking | Dec 2024 | Aug 2025 | Official leaderboard |
| 16 | | -0.41 | Non-thinking | Jul 2024 | Aug 2025 | Official leaderboard |
| 17 | | -1.56 | Non-thinking | Jul 2024 | Aug 2025 | Official leaderboard |
| 18 | | -2.56 | Non-thinking | Apr 2024 | Aug 2025 | Official leaderboard |
Rankings by part
We read 4 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Math Interaction, Math Outcome and Doc Interaction. GPT-5 leads 2 of them, and Claude 3.7 Sonnet leads the other 2. Arrows show how far a model moved from its overall rank among the models we track.
Math Interaction
18 models we track have a score here.
- 1 GPT-5 8.89
- 2 GPT-4.1 ↑5 8.87
- 3 GPT-4o Nov 2024 ↑2 8.84
- 4 Claude Sonnet 4 ↑2 8.74
- 5 Claude Opus 4.1 ↓2 8.71
- 6 Claude 3.7 Sonnet ↓4 8.70
- 7 Claude 3.5 Sonnet ↑2 8.66
- 8
Phi-4 8.66
- 9 GPT-4 Turbo Apr 2024 ↓5 8.60
- 10 GPT-4o mini 8.56
Math Outcome
18 models we track have a score here.
- 1 Claude 3.7 Sonnet ↑1 90%
- 2 GPT-5 ↓1 90%
- 3 GPT-4 Turbo Apr 2024 ↑1 84%
- 4
Phi-4 ↑4 84%
- 5 Claude Opus 4.1 ↓2 82%
- 6 Claude 3.5 Sonnet ↑3 76%
- 7 GPT-4o mini ↑3 76%
- 8 GPT-4o Nov 2024 ↓3 76%
- 9 GPT-4.1 ↓2 76%
- 10
Llama 3.1-70B ↑6 70%
Doc Interaction
18 models we track have a score here.
- 1 Claude 3.7 Sonnet ↑1 9.10
- 2 Claude Opus 4.1 ↑1 9.10
- 3 GPT-4.1 ↑4 9.08
- 4 GPT-5 ↓3 9.08
- 5 Claude Sonnet 4 ↑1 9.07
- 6 Claude 3.5 Sonnet ↑3 9.06
- 7 GPT-4 Turbo Apr 2024 ↓3 9.04
- 8
Gemini 2.5 Flash Jun 2025 ↑3 9.04
- 9 GPT-4o Nov 2024 ↓4 9.02
- 10
Gemini 2.5 Pro Jun 2025 ↑2 9.02
Doc Outcome
18 models we track have a score here.
- 1 GPT-5 8.96
- 2 Claude Opus 4.1 ↑1 8.90
- 3 Claude Sonnet 4 ↑3 8.80
- 4 Claude 3.7 Sonnet ↓2 8.73
- 5
Gemini 2.5 Flash Jun 2025 ↑6 8.70
- 6
Gemini 2.5 Pro Jun 2025 ↑6 8.66
- 7 GPT-4o Nov 2024 ↓2 8.59
- 8 GPT-4 Turbo Apr 2024 ↓4 8.50
- 9 GPT-4.1 ↓2 8.47
- 10 Claude 3.5 Sonnet ↓1 8.41
3 models have held number one on the official leaderboard since 2023
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official SimulatorArena leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official SimulatorArena 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 Instructions and chat
- IFEval In use Models answer about 500 prompts containing verifiable instructions such as word limits or required keywords. Code checks each instruction for strict and loose accuracy.
Gemma 4 31B IT 98.9%
- IFBench In use 58 new out-of-domain verifiable output constraints. Scored by programmatic constraint checks (strict or loose). Flint 89.2%
- MultiChallenge In use Multi-turn conversations testing instruction retention, inference memory, versioned editing and self-coherence. Graded by LLM judge with rubrics. GPT-6.1 Sol 82%
- Multi-IF In use Models follow verifiable instructions over three-turn conversations in eight languages, extending IFEval. Scored by instruction accuracy at each turn. LFM3-1.7B 89.1%
- AdvancedIF In use Models respond to over 1,600 prompts with complex, multi-turn, and system-level instructions. A judge model checks each response against expert-written rubrics.
MAI-Thinking-1 85%
- MM-IFEval In use Multimodal models follow image-grounded instructions with exact output constraints. Checks mix rule-based verification and a judge model. GPT-4o 64.6%
Data audit
8 of 9 checks pass, 1 warning. Look first at the Newest score check. The newest came 15 months ago, in August 2025, for GPT-5. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 2 pass 2 notes
- Pass Independent runs Scores from the official leaderboard (18 entries).
- Pass Models scored 18 models we track.
- Warning Newest score The newest came 15 months ago, in August 2025, for GPT-5.
- Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5 (#71).
- Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on simulatorarena.ai: 18 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.