Claude Code Pass Pricing: What's in the $9.99/Month Plan
Claude Code Pass costs $9.99 a month for boosted rate limits on Kimi K3.1 Pro, GLM 5.3, DeepSeek V4.1 Flash, and more open models.
What is Claude Code Pass and how much does it cost?
Claude Code Pass is a $9.99 per month subscription from the Claude Code team that gives you boosted API rate limits, roughly two to five times standard access, across a set of open models. Instead of paying per token to each model provider separately, you pay one flat fee and get higher throughput on models like Kimi K3.1 Pro, GLM 5.3, DeepSeek V4.1 Pro, DeepSeek V4.1 Flash, MiniMax M3.1 Flash, and Qwen 3.8 Max. It’s aimed at people running agentic coding workflows that burn through a lot of tokens, where per-token billing adds up fast.
TL;DR
- Claude Code Pass costs $9.99 a month and multiplies your API rate limits by two to five times compared to standard access on supported models.
- The plan bundles access to several open models, including Kimi K3.1 Pro, GLM 5.3, DeepSeek V4.1 Pro, DeepSeek V4.1 Flash, MiniMax M3.1 Flash, and Qwen 3.8 Max.
- Usage is capped with layered limits: a 5-hour window, a weekly cap, and a monthly cap, all visible on the Claude dashboard.
- It works across the Claude Code CLI, the browser/editor extension, and Claude Desktop, with sessions and history synced between all three.
- Claude Desktop adds features that pair well with the Pass, like importing a session that ran out of limits elsewhere and continuing it on a Pass-included model, and scheduling recurring jobs on a chosen model and workspace.
- For token-heavy workflows, such as design or research skills that can burn hundreds of thousands of tokens per run, a flat-rate plan changes the economics compared to metered API pricing.
- Third-party tools like Panda Code can plug into Claude Pass as a model provider, letting you assign different Pass-included models to different agents or roles in a larger workflow.
What exactly do you get for $9.99 a month?
The core value of Claude Code Pass is rate limits, not raw model access you couldn’t get elsewhere. The models bundled in are already available through their own APIs or other routers. What the Pass changes is how much you can throw at them before hitting a wall. Claude Code Pass advertises two to five times the rate limits of standard access on its supported models, which matters most if you’re running agents that make dozens or hundreds of calls per task.
The model lineup mentioned alongside the Pass includes Kimi K3.1 Pro, GLM 5.3, DeepSeek V4.1 Pro, DeepSeek V4.1 Flash, MiniMax M3.1 Flash, and Qwen 3.8 Max. These are mostly open-weight models from different labs, not Anthropic’s own Claude models. That’s worth noting: this isn’t a discount on Claude itself, it’s a flat-rate gateway to a handful of strong open alternatives with generous throughput.
Usage isn’t unlimited. The plan enforces caps on three time horizons: a rolling 5-hour window, a weekly total, and a monthly total. You can track where you stand against each of these on the Claude dashboard, which is useful if you’re running scheduled jobs or multiple parallel agents and want to avoid getting throttled mid-task.
Why does flat-rate pricing matter for agentic coding?
The pitch becomes clearest when you look at how token-hungry some agent workflows actually are. One example: a design automation skill (built to make AI-generated web pages look less generic by pulling from studied design patterns and real archival photography) was clocked at roughly 30 to 45 minutes per build and a few hundred thousand tokens consumed in that time. Run that on a model billed per token and costs scale linearly with every page you generate. Run it through Claude Pass and that cost is absorbed into the flat monthly fee instead.
The same logic shows up in multi-agent setups, where one lightweight “router” model hands off tasks to one or more heavier “worker” models that actually do the coding. These architectures can spin up many sessions and read a lot of cached context repeatedly. When token volume is high and recurring, a subscription with boosted limits is simply a better fit than metered billing, because the marginal cost of one more run doesn’t show up on an invoice.
This is also why cheaper or lighter models in the bundle, like DeepSeek V4.1 Flash, matter even if they’re not the most capable option. In workflows where a model is just making routing decisions or checking cached context rather than writing code, a fast cheap model covered by the Pass can be the better technical fit, not just the better financial one.
How does Claude Code Pass fit with the CLI, extension, and desktop app?
Claude Code Pass isn’t tied to a single interface. It works through the Claude Code CLI (the terminal-based agent), the browser or editor extension, and Claude Desktop, a standalone app version of the same agent harness that doesn’t require an editor at all. All three share the same underlying agent engine, and sessions are synced across them, so a conversation started in the CLI can be picked up later in the desktop app, with history visible in one place.
Two Claude Desktop features line up well with how the Pass is meant to be used. The first is session import: if you start a task in Claude Code and run out of limits partway through, you can import that conversation into Claude Desktop and continue it on a Pass-included model like Kimi K3.1 Pro instead of waiting for your original limits to reset. The second is scheduling: you can define a job with a name, frequency, prompt, model, and workspace, and have it run automatically. That’s a natural way to use Pass-covered models for routine maintenance work, like running a design or code checker over a project on a regular cadence without manually kicking it off each time.
Is Claude Code Pass worth it compared to paying per token?
Whether the $9.99 price makes sense depends entirely on your usage pattern. If you run occasional, lightweight agent tasks, metered API pricing on a single model might cost less overall since you’re not paying a fixed fee for headroom you don’t use. But if you’re running agentic workflows that are naturally token-heavy, multi-step design or coding skills, scheduled background jobs, or multi-agent systems with routers and workers, the math shifts quickly in favor of a flat plan.
The clearest signal from real-world use is that high-rate-limit flat plans make the cost of experimentation disappear. Running a build that burns a few hundred thousand tokens stops being something you have to think twice about. That matters more as agent workflows get more automated and less supervised, since the whole point of automation is running things repeatedly without re-evaluating the cost each time.
The tradeoff is that you’re locked into the specific models bundled with the plan. If your workflow depends on a model outside that list, Claude Code Pass doesn’t help you there, and you’d still need a separate API key or another provider.
Frequently Asked Questions
How much does Claude Code Pass cost?
Claude Code Pass is priced at $9.99 per month.
What models are included with Claude Code Pass?
The plan includes access to Kimi K3.1 Pro, GLM 5.3, DeepSeek V4.1 Pro, DeepSeek V4.1 Flash, MiniMax M3.1 Flash, and Qwen 3.8 Max, among others.
What are the rate limits on Claude Code Pass?
Claude Code Pass gives roughly two to five times the rate limits of standard access on supported models, with separate caps tracked over a 5-hour window, a week, and a month. Usage against each limit is visible on the Claude dashboard.
Does Claude Code Pass work with Claude Desktop?
Yes. Claude Desktop uses the same agent harness as the CLI and extension, and sessions sync across all three, so you can start a task in one and continue it in another.
Is Claude Code Pass cheaper than paying per token?
It depends on usage. For light, occasional tasks, per-token billing on a single model may cost less. For token-heavy or frequent agentic workflows, the flat monthly fee tends to work out cheaper and removes the need to track per-run costs.
