Deepseek V3.2 Thinking costs $0.620/1M input tokens and $1.85/1M output tokens as of 2026-06-27. DeepSeek publishes this model with a 128K tokens context window. o10 lists Deepseek V3.2 Thinking on DeepSeek at routing band C4. API slug deepseek-v3.2-thinking. Verify list prices against the venue you call.
Last updated: 2026-06-27. Full catalog. Reference prices are dated snapshots. Compare total cost and measured task quality before choosing a route.
Deepseek V3.2 Thinking. Price, context, and routing
Deepseek V3.2 Thinking is a chat model from deepseek (slug: deepseek-v3.2-thinking). Pick Deepseek V3.2 Thinking for high-volume DeepSeek chat at aggregator prices. It is the V-line, not R1 reasoning. Published input pricing starts at $0.620/1M tokens. Context window: 128K tokens.
Catalog snapshot 2026-06-27.
When should you pick Deepseek V3.2 Thinking?
Pick Deepseek V3.2 Thinking for high-volume DeepSeek chat at aggregator prices. It is the V-line, not R1 reasoning. Deepseek V3.2 Thinking is cheaper than most closed flagships. Escalate to DeepSeek R1 or a closed lab only when evals fail.
How does Deepseek V3.2 Thinking differ from sibling models?
Deepseek V3.2 Thinking is cheaper than most closed flagships. Escalate to DeepSeek R1 or a closed lab only when evals fail.
How much does Deepseek V3.2 Thinking cost per million tokens?
Deepseek V3.2 Thinking costs $0.620/1M input tokens and $1.85/1M output tokens in the gateway catalog snapshot (2026-06-27). Endpoint pricing may differ by fulfillment host.
Sourced from gateway catalog snapshot.
Where can you run Deepseek V3.2 Thinking?
Deepseek V3.2 Thinking is configured on DeepSeek. Known host IDs: deepseek: deepseek/deepseek-v3.2-thinking. This model is routable through o10 eval-gated routing.
What is Deepseek V3.2 Thinking's context window?
Deepseek V3.2 Thinking supports a 128K tokens context window in the gateway catalog snapshot. Context limits may vary by fulfillment host endpoint.
What o10 routing band is Deepseek V3.2 Thinking?
Deepseek V3.2 Thinking is classified at o10 routing band C4. Bands group models by cost tier and eval profile: C0 (economy), C1 (standard), C2 (capable), C3/C4 (frontier). Production routing is enabled for this slug.
Who makes Deepseek V3.2 Thinking?
Deepseek V3.2 Thinking is attributed to source provider deepseek in the o10 catalog. Fulfillment hosts (DeepSeek) supply API access; list pricing may differ by venue.
01Deep dive
Specifications
Deepseek V3.2 Thinking in the o10 unified catalog.
Deepseek V3.2 Thinking specifications
Field
Value
o10 slug
`deepseek-v3.2-thinking`
Type
text
Band
C4
Routable
Yes
Source provider
deepseek
Primary host
DeepSeek
Gateway ID
`deepseek/deepseek-v3.2-thinking`
Context window
128,000 tokens
Released
2025-12-01
02Deep dive
Token pricing
Gateway snapshot pricing for Deepseek V3.2 Thinking.
Deepseek V3.2 Thinking token pricing
Unit
Price
Input
$0.62/1M tokens
Output
$1.85/1M tokens
03Deep dive
Fulfillment hosts
Upstream venues configured for Deepseek V3.2 Thinking.
Deepseek V3.2 Thinking fulfillment routes
Host
Upstream model ID
deepseek
deepseek/deepseek-v3.2-thinking
04Deep dive
Gateway provider endpoints
Endpoint-level pricing and latency from the gateway catalog.
Deepseek V3.2 Thinking costs $0.620/1M input tokens and $1.85/1M output tokens in the gateway snapshot (2026-06-27). Verify against your provider's published pricing.
What is Deepseek V3.2 Thinking's context window?
Deepseek V3.2 Thinking supports 128K tokens in the gateway catalog snapshot. Host-specific endpoints may differ.
Which hosts serve Deepseek V3.2 Thinking?
Deepseek V3.2 Thinking (deepseek-v3.2-thinking) is configured on: DeepSeek. Upstream model IDs are listed in the fulfillment routes section.
Is Deepseek V3.2 Thinking routable through o10?
Yes. Deepseek V3.2 Thinking is routable at band C4. Send model: "deepseek-v3.2-thinking" or use an o10 smart routing alias.
How do I call Deepseek V3.2 Thinking via o10?
Use o10 slug "deepseek-v3.2-thinking" or gateway ID "deepseek/deepseek-v3.2-thinking" depending on your integration. Gateway-compatible endpoints accept OpenAI-style /v1 requests.
How does o10 route this model?
o10 routes Deepseek V3.2 Thinking only when per-use-case evals clear at your quality floor. Band C4 indicates cost tier. See /kyi for governance and /models/routing for smart aliases.