Updated daily at 05:00 UTC

Language

GPT-3.5 Turbo

By OpenAI, in the GPT-3.5 family.

#102Knowledge & reasoning

Released: 1 Mar 2023
Status: Deprecated
Retired: 23 Oct 2026
Context: 16.4K
Max output: 4.1K
Input per 1M tokens: $0.50
Output per 1M tokens: $1.50
Open weights: No
License: proprietary
Parameters: —
Knowledge cutoff: Sep 2021
Reasoning: No
Modalities: text → text

Category scores

Out of 100, from independent and third-party results only, by method category-2026-09-29.1.

Category scores
Category	Score	Rank	Coverage
Knowledge & reasoning	5.7	102	2 of 4
Human preference	26.8	Not ranked	1 of 5

Benchmark results

Every result we hold for this model, with who ran it and where it was published.

Knowledge & reasoning

Knowledge & reasoning results
Benchmark	Metric	Value	Provenance	Setting	Evaluated	Source
Adversarial NLI (ANLI)	Accuracy	58.1%	Self-reported	—	—	Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone via Epoch AI
AI2 Reasoning Challenge (ARC)	Challenge set accuracy	87.4%	Self-reported	—	—	Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone via Epoch AI
BIG-Bench Hard (BBH)	Average accuracy	61.6%	Third party	—	—	Baichuan 2: Open Large-scale Language Models via Epoch AI
Chess Puzzles	Puzzles solved	0.0%	Independent	—	7 Aug 2026	Epoch AI
DTBench (Decision Theory Benchmark)	Accuracy (capability questions)	48.5%	Third party	—	—	https://conceptualreasoning.ai/dtbench via Epoch AI
LMCA (Language Model Conceptual Argumentation)	Correlation with expert ratings (Pearson r)	0.0965	Third party	—	—	https://conceptualreasoning.ai/lmca via Epoch AI
MMLU	Accuracy	68.9%	Third party	—	—	Stanford CRFM Leaderboard via Epoch AI
MMLU	Accuracy	67.3%	Third party	—	—	Stanford CRFM Leaderboard via Epoch AI
MMLU	Accuracy	71.4%	Self-reported	—	—	Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone via Epoch AI
Mystery Game Puzzles	Puzzles solved	3.0%	Independent	—	27 Aug 2026	Epoch AI
OpenBookQA	Accuracy	86.0%	Self-reported	—	—	Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone via Epoch AI
WinoGrande	Accuracy	81.6%	Third party	—	—	The Falcon Series of Open Language Models via Epoch AI
WinoGrande	Accuracy	68.8%	Self-reported	—	—	Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone via Epoch AI

Scientific reasoning

Scientific reasoning results
Benchmark	Metric	Value	Provenance	Setting	Evaluated	Source
GPQA Diamond	Accuracy	28.0%	Independent	—	27 Jan 2025	Epoch AI
GPQA Diamond	Accuracy	27.2%	Independent	—	27 Jan 2025	Epoch AI

Mathematics

Mathematics results
Benchmark	Metric	Value	Provenance	Setting	Evaluated	Source
FrontierMath Tiers 1-3	Accuracy	0.0%	Independent	—	27 Aug 2026	Epoch AI
GSM8K	Exact match	57.8%	Third party	—	—	Baichuan 2: Open Large-scale Language Models via Epoch AI
GSM8K	Exact match	46.9%	Third party	—	—	Stanford HELM via Epoch AI
MATH Level 5	Accuracy	15.9%	Independent	—	27 Jan 2025	Epoch AI
MATH Level 5	Accuracy	11.6%	Independent	—	27 Jan 2025	Epoch AI
OTIS Mock AIME 2024-2025	Accuracy	2.2%	Independent	—	7 Aug 2026	Epoch AI

Coding & software engineering

Coding & software engineering results
Benchmark	Metric	Value	Provenance	Setting	Evaluated	Source
WeirdML (v2)	Average accuracy	3.5%	Third party	—	—	https://htihle.github.io/weirdml.html via Epoch AI

Factuality & hallucination

Factuality & hallucination results
Benchmark	Metric	Value	Provenance	Setting	Evaluated	Source
TriviaQA	Exact match	85.8%	Self-reported	—	—	Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone via Epoch AI

Human preference

Human preference results
Benchmark	Metric	Value	Provenance	Setting	Evaluated	Source
Text Arena (arena.ai)	Arena score (Elo scale)	1226	Independent	—	25 Sep 2026	Arena
Text Arena (arena.ai)	Arena score (Elo scale)	1204	Independent	—	25 Sep 2026	Arena
Text Arena (arena.ai)	Votes	66207	Independent	—	25 Sep 2026	Arena
Text Arena (arena.ai)	Votes	16619	Independent	—	25 Sep 2026	Arena

Other GPT-3.5 versions

Sources