LIVE
OPUS 4.7$15 / $75per Mtok
SONNET 4.6$3 / $15per Mtok
GPT-5.5$10 / $30per Mtok
GEMINI 3.1$3.50 / $10.50per Mtok
SWE-BENCHleader Claude Opus 4.772.1%
MMLU-PROleader Opus 4.788.4
VALS FINANCEleader Opus 4.764.4%
AFTAv1.0 whitepaper live at /whitepaper
OPUS 4.7$15 / $75per Mtok
SONNET 4.6$3 / $15per Mtok
GPT-5.5$10 / $30per Mtok
GEMINI 3.1$3.50 / $10.50per Mtok
SWE-BENCHleader Claude Opus 4.772.1%
MMLU-PROleader Opus 4.788.4
VALS FINANCEleader Opus 4.764.4%
AFTAv1.0 whitepaper live at /whitepaper
All systems operational0 AI providers monitored, polled every 2 minutes
Live status
All endpoints

Premium LLM-Ready OpenRouter Model

1 credit
GET /api/premium/clean/openrouter/{model_id}

The /api/premium/clean/openrouter/{model_id} endpoint pulls one model from the daily OpenRouter catalog snapshot and returns an LLM-ready fact card. Pricing is normalized to USD per million tokens (the universal agent convention) with a derived blended_5_to_1 mix capturing the typical 5:1 input:output ratio. Capability flags (tools, vision, structured_outputs, reasoning, response_format) are extracted from supported_parameters + modality so agents can boolean-filter. is_free_tier and created_iso are pre-computed.

When to use this endpoint

When an agent needs facts about one specific model from the OpenRouter catalog (367+ entries, no per-id filter on the upstream public endpoint). Saves the ~270KB ingestion required to find one entry; delivers ~500B per call. The compression headline is the most dramatic in the suite.

Parameters

NameInTypeDescription
model_id*pathstringOpenRouter catalog id with namespace, URL-encoded slash (e.g. anthropic%2Fclaude-haiku-4.5)e.g. anthropic/claude-haiku-4.5

* required

Example response

{
  "ok": true,
  "source_format": "openrouter_v1_models",
  "target_format": "tensorfeed_llm_ready_v1",
  "compression_stats": { "source_bytes": 274816, "cleaned_bytes": 542, "reduction_pct": 99.8, "approx_tokens_saved": 67429 },
  "data": {
    "id": "anthropic/claude-haiku-4.5",
    "name": "Anthropic Claude Haiku 4.5",
    "namespace": "anthropic",
    "context_length": 200000,
    "modality": "text+image->text",
    "pricing_per_million_tokens": { "prompt": 1, "completion": 5, "blended_5_to_1": 1.67, "image_per_image": null, "request_per_call": null },
    "capabilities": { "tools": true, "vision": true, "structured_outputs": true, "reasoning": false, "response_format": true },
    "max_output_tokens": 64000,
    "is_moderated": true,
    "is_free_tier": false,
    "created_iso": "2026-04-15"
  },
  "catalog_meta": { "captured_at": "2026-05-09T14:00:00Z", "total_models_in_catalog": 367 },
  "billing": { "credits_charged": 1, "credits_remaining": 49 }
}

Code samples

Python SDK

from tensorfeed import TensorFeed

tf = TensorFeed(token="tf_live_...")
m = tf._get("/premium/clean/openrouter/anthropic/claude-haiku-4.5")
p = m["data"]["pricing_per_million_tokens"]
print(f"${p['prompt']}/M input + ${p['completion']}/M output, {m['compression_stats']['approx_tokens_saved']} tokens saved")

TypeScript SDK

const res = await fetch(
  "https://tensorfeed.ai/api/premium/clean/openrouter/anthropic%2Fclaude-haiku-4.5",
  { headers: { Authorization: "Bearer tf_live_..." } }
);
const m = await res.json();
console.log(m.data.pricing_per_million_tokens.blended_5_to_1);

FAQ

Why is blended_5_to_1 useful?

Most agent workloads run a 5:1 input:output token ratio. Comparing models by prompt-only or completion-only price misses the real cost surface. blended_5_to_1 = prompt_per_million * (5/6) + completion_per_million * (1/6) and is what an agent should actually compare on.

What if the model id is not in the catalog?

Returns 404 with the requested id, the current catalog_size, and the captured_at timestamp echoed back so you can audit whether the model was removed in a recent snapshot or was never present. Use /api/openrouter/models to list the current catalog.

How fresh is the snapshot?

Captured daily at 14:00 UTC. The capturedAt field is included so agents can decide whether the freshness fits their use case. No staleness guarantee at the AFTA level (catalog moves slowly enough that 24h freshness is adequate).

Related endpoints