Appendix A — Providers, models & prices

This appendix is the thClaws.cloud gateway catalogue — the models you can call when you point an agent at https://thclaws.cloud/gateway (or run thClaws.cloud self-hosted). Desktop / CLI builds have their own catalogue covering 22 providers; see Chapter 6 — Providers, models & API keys for that list.

Catalogue refreshed 2026-06-02. To re-pull the latest models + rates from upstream APIs and LiteLLM’s pricing feed, run python3 scripts/refresh-model-catalogue.py from the repo root.

Pricing model

Rates below are what you pay — that is, upstream cost × platform markup. Internally the gateway keeps two pieces:

This means: if Anthropic raises Opus 4.8 to $6/M, you only update LiteLLM (or wait for the next sync) and the gateway picks up the new rate on the next pricing refresh — the markup stays a single env-var knob.

All prices are in US dollars per 1,000,000 tokens (1M tokens). The DB stores these as microcents per 1k tokens (µ¢/kt); the formula is $/M = µ¢/kt / 100,000.

Anthropic

Note — the tier column is display-only. thClaws.cloud dropped the tier-gating ladder in v0.28 — any user with positive credit can call any active model; per-call price differential is the only gate. See ch27 § “Why no tier gate” for the rationale.

OpenAI

OpenAI exposes ~76 chat-capable models via /v1/models (every gpt-5.x, o3, o4 variant, plus dated snapshots). Only the three above are currently seeded into the gateway. If you need a specific model, run the refresh script with --providers openai --apply to add it; it will pull pricing from LiteLLM automatically. The full list available upstream:

• gpt-5.5, gpt-5.5-pro, gpt-5.4, gpt-5.4-pro, gpt-5.4-mini, gpt-5.4-nano, gpt-5.3-codex, gpt-5.3-chat-latest
• gpt-5.2, gpt-5.2-pro, gpt-5.2-codex, gpt-5.2-chat-latest
• gpt-5.1, gpt-5.1-codex, gpt-5.1-codex-max, gpt-5.1-codex-mini
• gpt-5, gpt-5-pro, gpt-5-codex, gpt-5-mini, gpt-5-nano, gpt-5-search-api
• gpt-4.1, gpt-4.1-mini, gpt-4.1-nano
• o3, o3-pro, o3-mini, o3-deep-research, o4-mini, o4-mini-deep-research, o1-pro
• Legacy: gpt-4, gpt-4-turbo, gpt-3.5-turbo*

Google (Gemini)

Google exposes 20+ Gemini models via /v1beta/models (gemini-2.5-flash, gemini-2.5-pro, gemini-3-pro-preview, gemini-3.1-flash-lite, etc.). Run the refresh script to seed them with current pricing.

OpenRouter

The DB row for openrouter/auto holds 0/0 because the upstream cost varies per request. We rely on OpenRouter’s own metering in this case.

Inactive / deprecated rows

Remarks on accuracy

If you spot a price drift, run:

python3 scripts/refresh-model-catalogue.py --reprice-only            # dry-run
python3 scripts/refresh-model-catalogue.py --reprice-only --apply    # commit

Adjusting the platform markup

The 1.25× multiplier is a single env var on the gateway service:

# thclaws-cloud/docker-compose.yml
gateway:
  environment:
    THCLAWS_PLATFORM_MARKUP: ${THCLAWS_PLATFORM_MARKUP:-1.25}

Changing it to 1.50 and bouncing the gateway is the fastest way to widen margin uniformly across every model. The gateway clamps any value below 1.0 (running at exact upstream cost is the minimum — anything lower would mean losing money per call) and emits a warning.

For per-model differential pricing, edit the model_pricing row directly — but note that the refresh script will reprice it back to LiteLLM on the next --reprice run. Mark such rows by setting active=FALSE then active=TRUE with a custom rate, and exclude them from refresh with --providers scoping.

Where the rates live in code