sitdraw

Measured: Mistral Large 4 OpenRouter $0.70/$2.10 vs HF 12060 downloads

OpenRouter lists $0.70/$2.10 per million tokens and 1048576 context; the official Hub repo shows 12060 downloads and 1644 likes.

On 2026-11-10 the check is OpenRouter id mistralai/mistral-large-4-0: list price, 1048576 context, and downloads/likes on the Hugging Face repo that id points to.

Verify Mistral Large 4 OpenRouter price. List: https://openrouter.ai/api/v1/models. Hub search: https://huggingface.co/api/models?search=mistral-large-4. Only those two same-day responses. sitdraw · guide.

List price1048576 contextOfficial downloads

OpenRouter row, same day

FieldValue on 2026-11-10Note
Model idmistralai/mistral-large-4-0name: Mistral: Mistral Large 4
canonical_slugmistralai/mistral-large-4-0-20261006created 1791294122 → 2026-10-06 13:42:02 UTC
hugging_face_idmistralai/Mistral-Large-4-1T-A52BOfficial repo exists
context_length1048576Same on top_provider
Max completion262144is_moderated false
Input / 1M tokens0.0000007 USD → $0.70USD-per-token × 1,000,000
Output / 1M tokens0.0000021 USD → $2.10
Cache read / 1M tokens0.00000007 USD → $0.07
Modalitytext+image → textinputs text, image; output text
ReasoningNot mandatory; default on; efforts high / none; default highNo latency or homemade scores here

The description calls it a frontier multimodal (text and image input) model for reasoning, coding, and agentic workloads, and mentions a 1M-token window; the API string is cut at “with up”. knowledge_cutoff and expiration_date are empty.

Hugging Face: the official repo matches

The same-day search mistral-large-4 returned 68 rows. The table keeps the OpenRouter hugging_face_id, the org NVFP4 repo, and the highest-download community GGUF in separate cells. Downloads are not summed.

RepodownloadslikescreatedAtRelation
mistralai/Mistral-Large-4-1T-A52B1206016442026-09-30OpenRouter hugging_face_id; pipeline image-text-to-text; gated false; 69 siblings; lastModified 2026-11-02
mistralai/Mistral-Large-4-1T-A52B-NVFP41235712026-10-15Same-org quant; not the listed hugging_face_id
unsloth/Mistral-Large-4-1T-A52B-GGUF1771349(search hit)Third-party GGUF; more downloads than official; not a substitute
0.70input USD / 1M tokens
2.10output USD / 1M tokens
12060official downloads

Three sentences this measurement supports

List price is $0.70 input and $2.10 output per million tokens, with context_length 1048576. hugging_face_id points at mistralai/Mistral-Large-4-1T-A52B (12060 downloads, 1644 likes that day). Extra quant repos in search must stay separate; they are not latency or eval scores.

Check it yourself next time

  1. 01
    Pull the OpenRouter list

    Find mistralai/mistral-large-4-0. Read pricing, context_length, hugging_face_id. Multiply unit price by 1,000,000.

  2. 02
    Open the official repo

    Read downloads and likes on mistralai/Mistral-Large-4-1T-A52B. Do not add NVFP4 or GGUF into those cells.

  3. 03
    Treat search as a cross-check

    search=mistral-large-4 only confirms the card is still there. Higher community downloads mean another card was fetched more often.

# 2026-11-10
# OpenRouter id: mistralai/mistral-large-4-0
# created: 1791294122 (2026-10-06 13:42:02 UTC)
# context_length: 1048576 · max_completion: 262144
# prompt: 0.0000007 USD/token → $0.70 / 1,000,000
# completion: 0.0000021 USD/token → $2.10 / 1,000,000
# input_cache_read: 0.00000007 USD/token → $0.07 / 1,000,000
# hugging_face_id: mistralai/Mistral-Large-4-1T-A52B
# HF official: downloads 12060, likes 1644, createdAt 2026-09-30
# HF search mistral-large-4: 68 hits
How were $0.70 and $2.10 computed?

From OpenRouter pricing.prompt 0.0000007 and pricing.completion 0.0000021, each times 1,000,000. Not a screenshot of a bill, and not a latency test.

Does the official Hub repo exist?

Yes. hugging_face_id is mistralai/Mistral-Large-4-1T-A52B; on 2026-11-10 downloads 12060, likes 1644, gated false.

Why does the community GGUF show more downloads?

unsloth/Mistral-Large-4-1T-A52B-GGUF had 17713 downloads that day. That is a different card. It does not change the official counts or the list price.

Are there eval scores on this page?

No. The brief is list price, context, downloads, and likes. Extra fields on the list object are not treated as our own benchmarks.

Price vs downloads, one page

Lay $0.70 / $2.10 next to 12060, then leave. sitdraw.

Start the picture