DeepSeek V3.2
Chat model · DeepSeek ·
DeepSeek V3.2 is a chat model by DeepSeek. In Endless it costs 22.7 credits per 1M input tokens and 34 per 1M output tokens, on the Hobby plan and up.
DeepSeek V3.2 is DeepSeek's low-cost text model for coding and step-by-step reasoning, and its first to combine thinking with tool use in one turn.
- Price
- 22.7 credits per 1M input tokens
- 34 credits per 1M output tokens
- Availability
- Available on Hobby and above
- Context window
- 128K tokens
- Inputs
- Text
- Outputs
- Text
About DeepSeek V3.2
DeepSeek V3.2 is a text model from DeepSeek built for reasoning and agent work at a low cost. It uses DeepSeek Sparse Attention, which cuts the compute needed for long inputs while keeping quality steady. It was the first DeepSeek model to reason while calling tools, and it can work with or without a thinking phase.
Teams reach for it on everyday engineering jobs where cost adds up: explaining an unfamiliar function, writing SQL from a schema, fixing a failing test or walking through a math or logic problem one step at a time. With thinking off, it gives direct answers for quick questions.
DeepSeek V4.1 Flash is the newer DeepSeek option in Endless, with image input and a much larger context. Stay with V3.2 for text-only coding and reasoning, and switch models in the same workspace chat when a task needs screenshots or very long files.
Best for
- Explaining and refactoring unfamiliar code
- Writing SQL queries from a schema
- Step-by-step math and logic problems
- Tool-using agents on a tight budget
Questions about DeepSeek V3.2
- Does DeepSeek V3.2 have a thinking mode?
- Yes. DeepSeek V3.2 can reason step by step before answering, or skip that phase for quicker replies, and it can call tools in both modes.
- What is DeepSeek Sparse Attention?
- It is the attention mechanism in DeepSeek V3.2. It reduces the computation needed to process long inputs while keeping output quality close to full attention.
- Can DeepSeek V3.2 read images?
- No. DeepSeek V3.2 works with text only. For screenshots or charts, use DeepSeek V4.1 Flash, which understands images.
Related models
DeepSeek V4.1 Flash
by DeepSeek
Quick, low-cost reasoning that can also read images
12.2 / 48.6 credits per 1M tokens
GPT-6.1 Sol
by OpenAI
Near-flagship agentic coding and professional work at lower cost
162 / 810 credits per 1M tokens
Claude Opus 5.5
by Anthropic
Long agentic coding and research work over a 1M-token context
324 / 1620 credits per 1M tokens
Give your next idea a place to grow.
Start free, create a project, and bring your brand’s references together. Or book a demo to explore Endless with your team.