Every model,
on one page.
Pricing, capabilities, and provenance for every model — kept current, structured for cost modelling, and queryable. The registry behind Switchboard's routing.
Tracked per model
How does Linecard help?
Every model, every capability
Track what each model actually supports — tool calling, vision, reasoning tokens, prompt caching — not what the marketing page promises. Capability flags are tested, not declared.
See the matrixPricing that stays current
Provider prices change. Linecard tracks per-token rates and cache write/read multipliers for every model, with a timestamp on every update so you can model cost confidently.
View pricingProvenance without guessing
Each row records where the model is served from, what version is active, and which inference path Switchboard takes. No more wondering whether you got the cached version or the latest weights.
See provenanceOne source of truth for routing
Linecard is what Switchboard reads to choose a provider. If a model's in Linecard, it's routable — no separate gating step. Add a model once, use it everywhere.
See SwitchboardSample rows
What a Linecard row looks like
| Model | Input / 1M | Output / 1M | Cache (read / write) | Tools | Vision |
|---|---|---|---|---|---|
| anthropic/claude-sonnet-4-5 | $3.00 | $15.00 | 0.10× / 1.25× | ✓ | ✓ |
| openai/gpt-5 | $5.00 | $20.00 | 0.50× / — | ✓ | ✓ |
| google/gemini-2.5-pro | $1.25 | $10.00 | 0.25× / — | ✓ | ✓ |
| deepseek/deepseek-v4 | $0.27 | $1.10 | 0.10× / — | ✓ | — |
| meta/llama-4-405b | $0.90 | $0.90 | — | ✓ | — |
Indicative pricing for illustration. The live Linecard reflects the current rate every provider charges, with timestamped updates.
Linecard FAQs
Is Linecard the same as a model catalog?
Similar, but Linecard is opinionated about what gets tracked — pricing, capability flags, provenance, last-benchmarked. Catalogs typically just list names. Linecard is structured for cost modelling and capability-based routing.
How fresh is the pricing data?
Provider rates are updated within hours of changes. Every row has a `priced_at` timestamp; pricing older than 7 days surfaces a stale-data badge.
Can I query Linecard directly?
Yes. A read API is available — query by model id, capability, price range, or provider. Pairs with Switchboard so you can route based on the same data your cost model uses.
Is Linecard open source?
Not yet. The data shape is documented; if open access becomes the right move for distribution, we'll consider opening the data first and the platform separately.
The other half
One API for every model in the registry
Switchboard routes to anything Linecard tracks. OpenAI-compatible wire format, per-end-user attribution, one invoice.
Pick by capability, not by name
Linecard is how you reason about model choice — and how Switchboard actually routes.