Pricing
Pay per token.
Top up, create a key, and go.
| Model | Input$ per 1M tokens | Output$ per 1M tokens | Cache read$ per 1M tokens |
|---|---|---|---|
DeepSeek V4.1 FlashFeatured DeepSeek's latest flash model, built for fast agentic coding ⚡ A 552B MoE (8B active per token on prefill, 16B on decode) with native image understanding and a 1M context window. The default for new chats and CLI setups. | $0.15 | $0.60 | $0.028 |
Kimi K3 The first open 3T-class model, built for repository-scale agentic coding. 1M context, native vision. | $3.00 | $15.00 | $0.30 |
GLM 5.3 Z.ai's flagship coding model: the GLM 5.2 successor on the same base, with large gains on complex, long-horizon tasks and a 1M context window. Always thinks (dial low to max). | $1.40 | $4.40 | $0.26 |
Umans Flash The light, fast complement for the roles around the main coder: context gathering, scout subagents, summaries, quick edits. Umans Flash runs Qwen3.6 35B A3B today. It moves to DeepSeek V4.1 Flash, at its rates, before Nov 1, 2026. | $0.15 | $1.00 | $0.05 |
GLM 5.3 FlashDeprecated · discontinued Nov 1, 2026 Z.ai's fast coding model: a 320B MoE with 18B active parameters, the first natively multimodal GLM-5 series release (images and video), on a 1M context window. Always thinks (dial low to max). Replaced by DeepSeek V4.1 Flash. Umans Coder is an alias for this model today, at the same rates. It moves to DeepSeek V4.1 Flash before Nov 1, 2026. | $0.15 | $0.50 | $0.03 |
DeepSeek V4 FlashDeprecated · discontinued Nov 1, 2026 The cheapest model in the lineup, with a 1M context window for long agentic runs. Text only. Replaced by DeepSeek V4.1 Flash. | $0.14 | $0.28 | $0.028 |
USD per 1M tokens. Top up the wallet, create a key, and every request debits exactly what the model used.
Top up & startLabs experiments are free for seat holders while they run, and founding users have priority on seats. Watch the status page for openings and retirements.