AI Providers
These are the third-party AI services AITWIRE sends your Queries to, and the AI models it uses to evaluate the responses. Section 8B of the Terms of Service refers to this page. AITWIRE sends only the Queries you approve, and responses are subject to each provider's terms.
Services your Queries are sent to
Anthropic — Claude API
- Answers your Queries using the provider's own web search. Model:
claude-haiku-4-5-20251001. - Answers your Queries from the model's own knowledge. Model:
claude-haiku-4-5-20251001.
- Answers your Queries using the provider's own web search. Model:
Cloudflare — Workers AI
- Answers your Queries from the model's own knowledge. Model:
@cf/meta/llama-3.3-70b-instruct-fp8-fast.
- Answers your Queries from the model's own knowledge. Model:
Google — Gemini API
- Answers your Queries with Grounding with Google Search. Model:
gemini-2.5-flash. Results from Grounding with Google Search are for you and your authorised users: AITWIRE does not publish, or share with other customers, results or statistics derived from them (Terms Section 8B). - Answers your Queries from the model's own knowledge. Model:
gemini-2.5-flash.
- Answers your Queries with Grounding with Google Search. Model:
Mistral AI — Mistral API
- Answers your Queries from the model's own knowledge. Model:
mistral-small-latest.
- Answers your Queries from the model's own knowledge. Model:
OpenAI — OpenAI API
- Answers your Queries using the provider's own web search. Model:
gpt-4o-mini. - Answers your Queries from the model's own knowledge. Model:
gpt-4o-mini.
- Answers your Queries using the provider's own web search. Model:
Perplexity — Sonar API
- Answers your Queries through the provider's search product. Model:
sonar.
- Answers your Queries through the provider's search product. Model:
xAI — xAI API (Grok)
- Answers your Queries using the provider's own web search. Model:
grok-4.3. - Answers your Queries from the model's own knowledge. Model:
grok-4.3.
- Answers your Queries using the provider's own web search. Model:
Models that evaluate the responses
To measure a response, AITWIRE sends it, with the Query and the facts you approved, to these models.
Anthropic — Claude API
- Checks each response against your approved facts (semantic scoring). Model:
claude-haiku-4-5-20251001. - Judges a sample of responses for accuracy against your facts. Model:
chosen by AITWIRE's AI gateway; AITWIRE specifies claude-sonnet-4-6. - Grades responses for citations, sentiment, quality, competitors and recommendations. Model:
chosen by AITWIRE's AI gateway; AITWIRE specifies claude-sonnet-4-6. - Writes the narrative of your reports from your measurement results, which can quote short excerpts of responses. Model:
chosen by AITWIRE's AI gateway.
- Checks each response against your approved facts (semantic scoring). Model:
Cloudflare — Workers AI
- Checks a response against your approved facts only when the primary scorer is unavailable. Model:
@cf/qwen/qwen3-30b-a3b-fp8. - Grades a sample of past responses in a staff calibration sitting, as a check on the judge. Model:
@cf/meta/llama-3.3-70b-instruct-fp8-fast. Used only when AITWIRE staff calibrate how responses are evaluated.
- Checks a response against your approved facts only when the primary scorer is unavailable. Model:
This list is generated from AITWIRE's engine configuration. A service appears here only while AITWIRE can send Queries or responses to it, and appears automatically when it starts.
The running configuration could not be read just now, so this list shows AITWIRE's committed configuration.