Smaller, latency-optimised Cloudflare decision model that returns probability-scored classifications of text or images
Intern-Decision
New· 4Open decision models from InternLM that score each permitted answer rather than generating text
About Intern-Decision
Routing a support request, moderating a post or choosing the next step for an agent are decisions a language model usually answers with a paragraph. Intern-Decision, from InternLM, answers with numbers instead. InternLM is the open-model team behind the Intern series, and it uploaded three open-weight decision models to Hugging Face on 26 September 2026, at 0.8B, 2B and 4B, and put the training and inference code on GitHub.
How a request works
Each request carries a shared state made of text and as many as eight images, along with a schema listing named questions. A question asks the model to pick one option, answer yes or no, or give a score. In a single forward pass it outputs a probability for each permitted answer to every question. Because it never produces free text, it cannot step outside the options supplied, although it may favour the wrong one. The inference engine described on the model card rejects inputs over 8,192 tokens by default.
Built on Qwen3.5
All three are fine-tuned from Alibaba's Qwen3.5 at matching sizes, and the vision encoder was not trained further. The weights are under the Apache 2.0 licence, but Qwen's own licence still applies beside it and merits a read before commercial use. The GitHub repository holds the code but not the training data.
Where it sits
InternLM places it alongside TypeSafe's Jev and Interfaze's Lev, a family aimed at routing, triage, moderation and agent decisions in large volumes, hosted on a business's own hardware without a per-call fee.
Figures InternLM reports
By InternLM's account, the 4B model reaches an average of 90.02 per cent over seven test suites where Jev reaches 88.74, and it replies in roughly 44 milliseconds on a single RTX 4090 where Jev needs roughly 110. The smallest, 0.8B, averages 79.38 per cent. These are InternLM's own runs. The team published no announcement, and the one outside write-up found repeats its figures without testing them.
TrustList summary
Intern-Decision is a trio of open-weight models from InternLM, sized at 0.8B, 2B and 4B, built to score options rather than write text. Given a shared state, optional images and a list of typed questions, each model returns a probability for every allowed answer rather than writing text. The benchmark figures are the developer's own, and no independent check had been published when the sources were read.
What Intern-Decision offers
- Option scoring: returns probabilities for each permitted answer to every question in a single pass.
- Question types: single choice from a list, yes or no, and scores, all defined in a schema.
- Image input: accepts up to eight images alongside the shared state.
- Routing and triage: suited to routing, triage, moderation and agent decisions made at high volume.
- Self-hosting: runs on a business's own hardware, with training and inference code on GitHub.
Who it serves
Teams that make many automated decisions and want to run the model on their own hardware.
Key facts
- Developer: InternLM.
- Published: 26 September 2026 on Hugging Face.
- Base models: fine-tunes of Alibaba's Qwen3.5 models of the same sizes.
- Licence: Apache 2.0 for the weights, with Qwen's licence kept alongside, which a business should review.
- Input limit: inputs over 8,192 tokens are rejected by default.
Summary written by TrustList from the business's own published information, Nov 2026.
Benchmarks & AI stats
| Benchmark | Official | Community avg |
|---|---|---|
| No benchmark scores yet. Be the first to add one. | ||
“Official” values are editor-approved and feed the ranking. “Community avg” is the mean of member submissions (shown for transparency; it never affects the ranking until an editor approves a value).
Sign in to add a benchmark score for this model.
Trust Score
An earned signal from verification, reviews, awards, transparency and engagement - the vendor can't buy it.
Updated 9/28/2026
Request a demo or quote from Intern-Decision
Reviews
Write the first review of Intern-Decision
Used it? Your experience helps other buyers decide.
Write a reviewQuestions & answers
No questions yet. Be the first to ask about Intern-Decision.
Intern-Decision alternatives
Similar listings buyers compare against this one.
Microsoft AI's text-to-speech model for keeping one voice across 23 languages, with voice cloning from short audio
Google's open multimodal embedding model that maps text, images, audio and video into one vector space, under Apache 2.0
Anthropic's smallest and fastest Claude 5.5 model, available on its own API and three cloud platforms
Black Forest Labs' image generation and editing model with bounding-box layouts, multiple reference images and 4K output
Microsoft AI's high-volume, low-latency variant of MAI-Voice-2.1 for speech in 23 languages, aimed at voice agents
Perplexity's open-weight decision model that picks among given options for routing and checks in AI systems
Microsoft AI's real-time speech-to-text model for voice agents, transcribing 60 languages with continuous language detection
More on TrustList
Everything here links back to the same verified catalogue. Pick your next stop.
- Write a reviewHelp other buyers decide
- CompaniesAgencies, consultancies and IT service providers, ranked by verified reviews.
- ProductsSoftware and SaaS with pricing, features, integrations and alternatives.
- ArticlesGuides, comparisons and research from the editorial desk and the community.
- AwardsAnnual recognition decided by verified reviews and an independent jury.
- LaunchesNew products and releases, voted up by the community every day.
- RequestsBuyers describe what they need; vendors respond directly.
- PeopleReviewers, authors and makers with public profiles.
- ComparePut up to four listings side by side before you shortlist.