DeepSeek V3.2 pricing
DeepSeek-V3.2 is a large language model designed to harmonize high computational efficiency with strong reasoning and agentic tool-use performance. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism that reduces training and inference cost while preserving quality in long-context scenarios. A scalable reinforcement learning post-training framework further improves reasoning, with reported performance in the GPT-5 class, and the model has demonstrated gold-me. Live index: 9 priced offers. Best input $0.260 per million tokens from DeepInfra. Best output $0.380 per million tokens from DeepInfra.
Pricing across providers
Every row is a seller of DeepSeek V3.2 with token pricing we track. The cheapest input in this snapshot is from DeepInfra. The bar chart shows the same input and output dollars per million for a quick scan.
| Provider | Input / 1M | Output / 1M | Cached input | Batch |
|---|---|---|---|---|
AB Aws Bedrock | $0.740 | $2.22 | — | — |
A Azure | $0.580 | $1.68 | — | — |
D DeepInfra | $0.260 | $0.380 | $0.130 | — |
F Friendliai | $0.500 | $1.50 | $0.250 | — |
G Gmi | $0.280 | $0.400 | — | — |
N Novita | $0.269 | $0.400 | $0.135 | — |
O Openrouter | $0.269 | $0.400 | $0.135 | — |
S Sambanova | $3.00 | $4.50 | — | — |
D Deepseeknative | $0.280 | $0.400 | $0.028 | — |
Input vs output · per provider
Pricing history
These entries are only for the primary native listing of DeepSeek V3.2, not every reseller row above. The native listing moved from $0.280 in and $0.400 out to $0.280 in and $0.400 out in the latest window.
Since Sep 21, 2026
$0.280 in · $0.400 out(current listing)
Mar 13, 2026 → Sep 21, 2026
$0.280 in · $0.400 out
Cost calculator
Pick any of the providers above and type how many tokens you expect per day, week, or year. We turn that into rough dollar totals for DeepSeek V3.2.
Provider
0.074000¢ / req
0.111000¢ / req
Model specifications
Context length, caps, and capability flags for DeepSeek V3.2. Family: DeepSeek V3. Values follow the main provider (Deepseek) record in our index.
- Context window
- 163,840 tokens
- Max output
- 163,840 tokens
- Vision (images)
- No
- Tool / function calling
- Yes
- Streaming
- No
- Released
- Dec 2025
- Primary provider
- Deepseek
- Model family
- DeepSeek V3
Compare DeepSeek V3.2
Open a pair page to see DeepSeek V3.2 next to another model with a shared provider matrix. 6 shortcuts below.
Locked
Compare with
Pick a model on both sides.
Popular DeepSeek V3.2 comparisons
- DeepSeek V3.2 vs GPT-4o
DeepSeek V3.2 96% cheaper on output
- DeepSeek V3.2 vs Claude Sonnet 4.6
DeepSeek V3.2 97% cheaper on output
- DeepSeek V3.2 vs Gemini 2.0 Flash
Same output pricing
- DeepSeek V3.2 vs Llama 3.1 70B
Compare pricing side by side
- DeepSeek V3.2 vs Mistral 7B
Mistral 7B 37% cheaper on output
- DeepSeek V3.2 vs GLM 4.7
DeepSeek V3.2 81% cheaper on output
Frequently asked questions
Answers pull from the same numbers you see on this page. The short model note from our index: DeepSeek-V3.2 is a large language model designed to harmonize high computational efficiency with strong reasoning and agentic tool-use performance. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse att...
Related models
Models close to DeepSeek V3.2 by family, provider, and context window — each with live pricing in our catalog.