DeepSeek-V3.1: Stable version, excellent long text
DeepSeek-V3.1 is part of the DeepSeek lineup from DeepSeek, designed for production workloads that need reliable quality at scale. Stable version, excellent long text.
It accepts Text inputs and produces Text outputs, with a context window of up to 163,840 input / 32,768 output tokens. The model is tuned for chat, analysis, and tool use in enterprise pipelines.
Balanced pre-training and alignment deliver solid instruction following, summarization, and structured JSON output.
Benchmarks
Benchmark (Area)
Score*
AIME 2024 (Math)
85.1% accuracy (Max)
LiveCodeBench (Coding)
80.6 Pass@1 (Max)
Codeforces Rating (Coding)
2685 rating (Max)
SWE-bench Verified (Coding)
69.5% resolved (Max)
MMLU-Pro (Knowledge)
75.9 EM (Max)
HMMT 2024 Feb (Math)
83.4 Pass@1 (Max)
GPQA Diamond (Science)
77.5 Pass@1 (Max)
MRCR 1M (Long Context)
69.3 MMR (Max)
Terminal Bench 2.0 (Agent)
50.1 Acc (Max)
Apex Shortlist (Agent)
75.4 Pass@1 (Max)
Chinese-SimpleQA (Chinese)
69.4 Pass@1 (Max)
SuperCLUE (Chinese)
60.56 overall (Max)
API request example
Endpoint /v1/chat/completions. Replace <API_KEY> with a key from the console.
Standard OpenAI Chat Completions. Swap `model` for the target slug.