DeepSeek

DeepSeek V3.2

164K ctxText Tools Reasoning

Released Sep 29, 2025 · 573B tokens this week · 5 providers

Overview

A sparse-attention update to the V3 line that cut long-context serving costs dramatically. It folds chat and reasoning modes into one checkpoint and is among the best value-for-quality models available. A frequent choice for cost-sensitive production traffic.

reasoningprogrammingagentsacademia

Providers

Reference data — not yet measured from live traffic

PROVIDER		MAX OUT		OUTPUT /M				QUANT
DeepSeekCheapest	164K	66K	$0.27	$0.40	104ms	58	99.40%	Full
Novita AI	164K	66K	$0.29	$0.42	640ms	88	98.60%	fp8
Together AI	164K	66K	$0.30	$0.45	195ms	105	99.50%	Full
Fireworks AI	164K	66K	$0.31	$0.46	520ms	128	99.55%	Full
SambaNova	66K	66K	$0.32	$0.47	380ms	240	99.20%	Full

Latency p50 reflects a provider’s API endpoint responsiveness — the round-trip to its API, not per-token inference time. These figures, with throughput and uptime, are reference data until the gateway aggregates its own traffic.

Price comparison

Blended $ per 1M tokens

DeepSeek V3.2$0.37

Grok 4 Fast$0.42

Gemini 3 Flash$1.95

GPT-5.2$10.94

Claude Sonnet 4.5$12.00

Blended rate per 1M tokens, weighted one part prompt to three parts completion — roughly the shape of a chat workload. Your mix will move the number.

Call it

OpenAI-compatible — swap the base URL and go

curl https://model.cards/api/v1/chat/completions \
  -H "Authorization: Bearer $MODELCARDS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek/deepseek-v3.2",
    "messages": [
      { "role": "user", "content": "Summarize the tradeoffs of speculative decoding." }
    ]
  }'