Catwalk v3 is here 💅

A brand-new, cleaner schema for describing models. Pricing, reasoning and capabilities each get their own home, and type now tells you how to talk to a model rather than who made it. Same catwalk, much better outfits.

The big one!

Provider type now names the API format a provider speaks, not the vendor. There are just three:

completions: OpenAI Chat Completions
responses: OpenAI Responses
messages: Anthropic Messages

Models can override their provider's type, too, so a single provider can serve Completions, Responses and Messages models side by side (looking at you, OpenCode and Copilot 👀).

Prettier models

Pricing now lives in a nested pricing object with clearer names: input, output, cache_create and cache_hit. No more guessing which "cached" field means what.
Reasoning is described by thinking (always, never or toggleable), plus effort levels that carry display names, like xhigh showing as "X-High".
Capabilities tell you about vision and, new, audio support.
max_attachments sets a per-model attachment limit.
session_affinity_header lets providers that route by session do just that.

Migrating from v2

The module path changed, so update your imports:

go get charm.land/catwalk/v3

import "charm.land/catwalk/v3/pkg/catwalk"

Then:

v2	v3
`TypeOpenAI`, `TypeAnthropic`, `TypeBedrock`…	`TypeCompletions`, `TypeResponses`, `TypeMessages` (use the provider `ID` to tell vendors apart)
`CostPer1MIn` / `CostPer1MOut`	`Pricing.Input` / `Pricing.Output`
`CostPer1MInCached`	`Pricing.CacheCreate`
`CostPer1MOutCached`	`Pricing.CacheHit`
`CanReason` (field)	`CanReason()` (method)
`ReasoningLevels`	`ReasoningEffortLevels()`, or `Reasoning.EffortLevels` for display names
`DefaultReasoningEffort`	`Reasoning.DefaultEffortLevel`
`SupportsImages`	`Capabilities.Vision`
—	`model.EffectiveType(provider.Type)` to resolve the per-model type

The client now fetches from /v3/providers.

Not ready to move yet? No worries. /v2/providers stays live for some time, serving the old schema translated automatically from the v3 data, so existing clients keep working. We'll give plenty of notice before it goes away.

More to come, Charm XX ⭐

Changelog

Fixed

79e2a975e2ad55490d68cdb2ee262d57003622c2: fix: specify the project name in goreleaser (@andrinoff)
31c845e782f3f17ac9db1b90f1b508ffe18b15b4: fix: wire types in openai and minimax (@andrinoff)

Other stuff

6e154dee53ab2aca60002f27446165c886b2410e: catwalk v3 (#611) (@andreynering)
f241ec3b011880aa78a4874931b97e32f2a15c10: chore: auto-update generated files (#711) (@github-actions[bot])

Verifying the artifacts

First, download the checksums.txt file and the checksums.txt.sigstore.json file files, for example, with wget:

wget 'https://github.com/charmbracelet/catwalk/releases/download/v3.0.0/checksums.txt'
wget 'https://github.com/charmbracelet/catwalk/releases/download/v3.0.0/checksums.txt.sigstore.json'

Then, verify it using cosign:

cosign verify-blob \
  --certificate-identity 'https://github.com/charmbracelet/meta/.github/workflows/goreleaser.yml@refs/heads/main' \
  --certificate-oidc-issuer 'https://token.actions.githubusercontent.com' \
  --bundle 'checksums.txt.sigstore.json' \
  ./checksums.txt

If the output is Verified OK, you can safely use it to verify the checksums of other artifacts you downloaded from the release using sha256sum:

sha256sum --ignore-missing -c checksums.txt

Done! You artifacts are now verified!

Thoughts? Questions? We love hearing from you. Feel free to reach out on X, Discord, Slack, The Fediverse, Bluesky.