DeepSeek V3

Tool callingStructured output

Hugging Face Website

DeepSeek chat model for instruction following, coding, and analysis

Context: 164K
Released: 10mo ago
Parameters: 671B
Input / 1M tokens: $0.32
Output / 1M tokens: $0.89

Sources:models.devOpenRouter

Benchmarks

DeepSWE is the primary coding benchmark where an exact model evaluation exists. Other sourced evaluations appear below; missing scores are not zero.

Not evaluated in DeepSWE 1.1 as of 2026-09-22. Results are only shown for exact model versions in the official dataset. Official leaderboard.

Other published benchmarks

Evaluation	Reported score and source
GPQA diamond	71.2% Epoch AI · deepseek-chat Epoch AI evaluation
OTIS Mock AIME 2024-2025	48.9% Epoch AI · deepseek-chat Epoch AI evaluation

Benchmark compilation: Epoch AI, Capabilities & benchmarking (CC BY attribution). Scores are converted to percentages and model names matched to the Vertical Tabs catalog. Individual evaluations are credited to their original sources. External results may carry separate licences and evaluation configurations; do not interpret scores from different harnesses as directly comparable.

Specifications

Developer: DeepSeek
Model family: DeepSeek
Type: Language model
Predecessor: —
Successor: —
Released: December 1, 2025
Model ID: deepseek/deepseek-chat
Openness: Open weights
License: —
Parameters: 671B
Input modalities: text
Output modalities: text
Modality: text->text
Native context: 160K
Tokenizer: DeepSeek
Max output tokens: 16K
Knowledge cutoff: 2025-09
Hugging Face: deepseek-ai/DeepSeek-V3
Source freshness: Metadata synced Oct 8, 2026, 10:17 PM UTC

Open-weight builds

Ollama variants

Every published Ollama tag matched to this model, including parameter editions, instruction/base variants, and quantized builds.

3 builds

Variant	Parameters	Type	Quantization	Size	Context	Input
deepseek-v3:671b-fp16	671B	—	FP16	1.3 TB	4K	text
deepseek-v3:latest	671B	—	Q4_K_M	404 GB	160K	text
deepseek-v3:671b-q8_0	671B	—	Q8_0	713 GB	4K	text

Additional source evidence

Source model: deepseek/deepseek-chat
Source family: deepseek
Last updated: 2026-02-28
Knowledge cutoff: 2025-09
Model context: 1M
Model output limit: 384K
Open weights: Yes
File attachments: Yes
Reasoning: No
Tool calling: Yes
Temperature control: Yes

Downloadable weights

Hugging Face

Published benchmarks

Benchmark	Score	Evaluation details	Evidence
Aider Polyglot	70.2 · percent correct	2025-10-03	Source

Complete source record (JSON)

{
  "id": "deepseek/deepseek-chat",
  "name": "DeepSeek Chat",
  "description": "DeepSeek chat model for instruction following, coding, and analysis",
  "family": "deepseek",
  "last_updated": "2026-02-28",
  "knowledge": "2025-09",
  "attachment": true,
  "reasoning": false,
  "tool_call": true,
  "temperature": true,
  "modalities": {
    "input": [
      "text"
    ],
    "output": [
      "text"
    ]
  },
  "limit": {
    "context": 1000000,
    "output": 384000
  },
  "benchmarks": [
    {
      "name": "Aider Polyglot",
      "score": 70.2,
      "source": "https://aider.chat/docs/leaderboards/",
      "metric": "percent correct",
      "date": "2025-10-03"
    }
  ],
  "release_date": "2025-12-01",
  "open_weights": true,
  "weights": [
    {
      "label": "Hugging Face",
      "url": "https://huggingface.co/deepseek-ai/DeepSeek-V3.2"
    }
  ]
}

Infrastructure

Provider availability

All providers

Provider	Endpoint	Context	Input /M tokens	Output /M tokens	Cached input /M tokens	Cache write /M tokens	Uptime 1d	Data handling
StreamLake	streamlake	128K	$0.26	$1.03	—	—	99.65%	ZDR: No Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15
DeepInfra	deepinfra/fp4	164K	$0.32	$0.89	—	—	87.64%	ZDR: Yes Training: Unknown Prompt logging: Unknown Retention: Unknown ZDR source Privacy policy Checked 2026-10-15

Provider data:OpenRouter

Deployment graph

Can I run this?

Vertical Tabs only shows compatibility when it has structured evidence tying this exact model to hardware, runtime, quantization, memory use, and observed performance.

Browse hardware