Model
Model explorer

GPT-3.5 Turbo

CLOSED
OpenAI · GPT-3.5 family · released Mar 1, 2023

Chat-tuned model behind ChatGPT's Nov 2022 launch; still served today with a 16K context window at $0.50/$1.50 per million tokens.

ReasoningCodingVisionFunction callingTool useAgentic
502.6
Elo · rank #303
Parameters
Undisclosed
Active params
—
Context
16K tokens
Architecture
Transformer (architecture undisclosed)
License
Proprietary
Languages
—
API price (in/out)
$0.5 / $1.5
Modalities
text
Benchmark results
Bar shows position within the tracked field; marker = field best
AlpacaEvalHuman preference22.4%#36
best: Tülu 2+DPO 70B · 95.1%
ARC-ChallengeReasoning85.2%#31
best: Llama 3.1 405B · 96.9%
Arena-HardHuman preference24.8%#52
best: Qwen3-235B-A22B (Non-Thinking) · 96.1%
Arena EloHuman preference1112#46
best: Claude Fable 5 · 1505
DROPReasoning64.1#48
best: Hunyuan-T1 · 93.1
GSM8KMath57.1%#120
best: Llama 3.1 405B · 96.8%
HellaSwagReasoning85.5%#37
best: Claude 3 Opus · 95.4%
APIBench (HuggingFace)Agents16.8%#3
best: Gorilla 7B · 71.7%
HumanEvalCoding48.1%#127
best: Claude Opus 4.5 · 99.4%
JASTERKnowledge56.7%#2
best: PLaMo-100B · 77.5%
LogicKorHuman preference7.76#17
best: GPT-4o · 9.33
MMLUKnowledge70.0%#145
best: OpenAI o3 · 92.9%
MT-BenchHuman preference7.94#48
best: Hunyuan-Large (A52B) · 9.4
OpenAssistant BenchHuman preference1015#2
best: GPT-4 · 1294
APIBench (TensorHub)Agents41.8%#2
best: Gorilla 7B · 83.8%
APIBench (TorchHub)Agents48.4%#2
best: Gorilla 7B · 59.1%
Vicuna BenchHuman preference916#5
best: GPT-4 · 1176
WinoGrandeReasoning81.6%#25
best: PaLM 2 · 90.9%
Run it locally
Closed weights — available via API only. No local deployment.
Input / M tok
$0.5
Output / M tok
$1.5
GPT-3.5 family
Elo progression across releases
GPT-3.5 TurboMar 2023502.6
Resources
DOCSdevelopers.openai.com/api/docs/models/gpt-3.5-turboAPIAvailable via OpenAI API
API price $0.5/$1.5 · each benchmark row carries its own source badge (see methodology)