795fadc0bc
To support the Agentic Editing launch. To dos before merging:
- [ ] Anything marked as `todo!` within `docs/src` (Anyone)
- [x] Check all internal links (Joe)
- Joe: I checked all links and fixed all aside from a few that I
annotated with `todo!` comments
- [ ] Update images (Danilo)
- [ ] Go over / show images of tool cards in agent panel overview
(Danilo)
- [ ] Point billing FAQ to new billing docs (Joe)
- [x] Redirects external links
- [ ] Needs testing
- [x] Delete old docs
- [ ] Ensure all mentioned bindings use the `{#kb ...}` format and that
they are rendering correctly
- [ ] All agent-related actions are now `agent::` and not `assistant::`
- [x] Mention support of `.rules` files in `rules.md`
Release Notes:
- N/A
---------
Co-authored-by: Joseph T. Lyons <josephtlyons@gmail.com>
Co-authored-by: morgankrey <morgankrey@gmail.com>
Co-authored-by: Smit Barmase <37347831+smitbarmase@users.noreply.github.com>
Co-authored-by: Ben Kunkle <Ben.kunkle@gmail.com>
Co-authored-by: Peter Tripp <peter@zed.dev>
Co-authored-by: Ben Kunkle <ben@zed.dev>
Co-authored-by: Mikayla Maki <mikayla.c.maki@gmail.com>
Co-authored-by: Ben Brandt <benjamin.j.brandt@gmail.com>
3.4 KiB
Models
Zed’s plans offer hosted versions of major LLM’s, generally with higher rate limits than individual API keys. We’re working hard to expand the models supported by Zed’s subscription offerings, so please check back often.
| Model | Provider | Max Mode | Context Window | Price per Prompt | Price per Request |
|---|---|---|---|---|---|
| Claude 3.5 Sonnet | Anthropic | ❌ | 120k | $0.04 | N/A |
| Claude 3.7 Sonnet | Anthropic | ❌ | 120k | $0.04 | N/A |
| Claude 3.7 Sonnet | Anthropic | ✅ | 200k | N/A | $0.05 |
Usage
The models above can be used with the prompts included in your plan. For models not marked with “Max Mode”, each prompt is counted against the monthly limit of your plan.
If you’ve exceeded your limit for the month, and are on a paid plan, you can enable usage-based pricing to continue using models for the rest of the month. See Plans and Usage for more information.
Non-Max Mode will use up to 25 tool calls per one prompt. If your prompt extends beyond 25 tool calls, Zed will ask if you’d like to continue which will consume a second prompt. See Max Mode for more information on tool calls in Max Mode.
Max Mode
In Max Mode, we enable models to use large context windows, unlimited tool calls, and other capabilities for expanded reasoning, to allow an unfettered agentic experience. Because of the increased cost to Zed, each subsequent request beyond the initial user prompt in Max Mode models is counted as a prompt for metering. In addition, usage-based pricing per request is slightly more expensive for Max Mode models than usage-based pricing per prompt for regular models.
Note that the Agent Panel using a Max Mode model may consume a good bit of your monthly prompt capacity, if many tool calls are used. We encourage you to think through what model is best for your needs before leaving the Agent Panel to work.
By default, all Agent threads start in normal mode, however you can use the agent setting preferred_completion_mode to start new Agent threads in max mode.
Context Windows
A context window is the maximum span of text and code an LLM can consider at once, including both the input prompt and output generated by the model.
In Max Mode, we increase context window size to allow models to have enhanced reasoning capabilities.
Each Agent thread in Zed maintains its own context window. The more prompts, attached files, and responses included in a session, the larger the context window grows.
For best results, it’s recommended you take a purpose-based approach to Agent thread management, starting a new thread for each unique task.
Tool Calls
Models can use tools to interface with your code, search the web, and perform other useful functions. In Max Mode, models can use an unlimited number of tools per prompt, with each tool call counting as a prompt for metering purposes. For non-Max Mode models, you'll need to interact with the model every 25 tool calls to continue, at which point a new prompt will be counted against your plan limit.