Pick a plan for private coding-agent inference, then scale with usage packs or a team plan as usage grows.
For individual developers & engineers
~300M tokens expected monthly usage
For high-frequency multi-agent workflows
~1.5B tokens expected monthly usage
For teams, orgs, and compliance requirements
Custom pooled usage plan
Agent web search is billed per search query. Available across all plans. Prompts and completions are never sent to Brave Search—only the standalone search query is transferred.
Direct pay-as-you-go rates across input, output, and cache read tokens.
| Model | Context | Input / 1M | Output / 1M | Cache Read / 1M |
|---|---|---|---|---|
GLM-5.2(Z.ai) | 524K | $1.10 | $4.00 | $0.20 |
DeepSeek V4 Flash 0731(DeepSeek) | 1M | $0.14 | $0.28 | $0.0028 |
Kimi K3(Moonshot AI) | 1M | $2.50 | $12.00 | $0.45 |
Laguna 2.1(NullStack) | 2M | Coming soon | — | — |
Add usage at any time. Usage packs never change your plan terms, and remain valid for 90 days.
Instant top-up for light overflow testing.
Standard extension for agent refactors.
Popular pack for extensive multi-day runs.
Heavy workload booster for repo migrations.
Maximum pack for autonomous agent fleets.
Estimate monthly costs with prefix caching discounts for coding sessions.
Coding agents typically achieve 40% - 75% prefix cache hits on large multi-file prompts.