The story so far
Weights, every dated article, and repo commits on one rail.
5 repo commits since the weights landed
Teal is the weights (2026-09-26), grey ticks are repo commits.
Intern-Decision-4B
internlm Apache 2.0 out 29 days ago
4.54B parameters, images and text in, text out, 256K tokens of context.
Verified 2026-10-25 Rank #1935 in Decision models · score 60.3 method v1 Sources 3 Downloads 30d 5.2k · #539 of 619 tracked
Do people use it?5.2kdownloads, 30 days#539 of 619 we track · below the median+3 likes, downloads +22% this week90 likesHugging Face2026-10-24
Where it's used
Hugging Face does not say who downloads a model or with what. These are the proxies: API providers listing it, quantized builds, fine-tunes/adapters/merges, and Spaces that load it.
No API provider serves it, so its downloads are people running it themselves: 3 quantized builds and 9 Spaces.
- As a service0no API provider listed
- On own hardware3quantized builds
- Built on43 fine-tunes · 1 adapters · 0 merges
- In apps9Hugging Face Spaces
Top Spaces running it: multimodalart/jev-decision-index ♥ 802 · hugging-apps/internlm-intern-decision-4b ♥ 1 · mertkayacs/JevAlt ♥ 1
Will it fit?
Weights only, an estimate: leave room for the KV cache. Method
Fits a 16 GB card at INT4
About 2.7 GB at INT4, 5.4 GB at FP8, 11 GB at BF16: weights only, so leave room for the KV cache.
Estimate params × bytes per param × 1.2 · how we estimate
What people built on it
7 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.
7 derivatives on the Hub, led by 3 quantized
From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.
Reference
Everything else, folded: the full tables, every source.
Benchmarks
14 rows
Benchmarks 14 rows
measured third-party leaderboard rowauthor-reported number from the model card or eval files (14)No confidence intervals unless the source gave them; settings differ between rows, so compare within a benchmark, not across. Date = the benchmark's own date when it gives one, else when we read it.
| Benchmark | Value | Setting / notes | Reported by | Date |
|---|---|---|---|---|
| author-reported AG News | 90.82 | Intern-Decision-4B | model author | 2026-10-23 |
| author-reported Average | 90.02 | Intern-Decision-4B | model author | 2026-10-23 |
| author-reported Brier ↓ | 0.35 | Intern-Decision-4B | model author | 2026-10-23 |
| author-reported ECE ↓ | 0.07 | Intern-Decision-4B | model author | 2026-10-23 |
| author-reported Jevbench-Easy | 100 | Intern-Decision-4B | model author | 2026-10-23 |
| author-reported Jevbench-Hard | 73.87 | Intern-Decision-4B | model author | 2026-10-23 |
| author-reported Jevbench-Original | 98.61 | Intern-Decision-4B | model author | 2026-10-23 |
| author-reported Mean | 44.16 ms other | Intern-Decision-4B latency | model author | 2026-10-23 |
| author-reported Median / P50 | 44.03 ms other | Intern-Decision-4B latency | model author | 2026-10-23 |
| author-reported Overall (Brier / ECE) | 0.550 / 0.089 | Intern-Decision-4B after calibration | model author | 2026-10-23 |
| author-reported P95 | 44.60 ms other | Intern-Decision-4B latency | model author | 2026-10-23 |
| author-reported ToolACE | 96.45 | Intern-Decision-4B | model author | 2026-10-23 |
| author-reported Typed Decision | 80.55 | Intern-Decision-4B | model author | 2026-10-23 |
| author-reported WildJailBreak | 89.86 | Intern-Decision-4B | model author | 2026-10-23 |
What the card says
2 cited sentences
Summary generated, cited
Intern-Decision-4B is a multimodal structured decision model with 4.54B parameters.[1][2] Fine-tuned from Qwen3.5-4B, the model processes images and text to return an answer distribution for questions in a single forward pass.[3][4]
- “multimodal structured decision model”
- parameters human
- “returns an answer distribution for every question in one …”
- base models
Generated 2026-10-07 15:17 UTC by parasail-gemma-4-26b-a4b-it; each numeral points at the fact it rests on. Not a verdict.
Use-case evidence
7 use cases
What it is good at 7 use cases
Labels are evidence, not verdicts: measured a third-party leaderboard row for this model; author-reported a number the model author published in the card or eval files; structural a deterministic signal from the name, config or chat template, or a decision-model estimate. Each chip links to its source. All use cases.
- Understand images and videostructuraldecision-model estimate p=0.82tag: multimodalpipeline: image-text-to-text+1 more
- Classify and make typed decisionsstructuraldecision-model estimate p=1.00name: Decisiontag: decision-model+1 more
- Run on a laptop or phonestructuraldecision-model estimate p=0.96name: 4Bfits: single-gpu
- Chat and follow instructionsstructuraltag: conversationalchat template present
- Call tools and APIsstructuralchat template: <tool_call>
- Reason through hard problemsstructuralchat template: <think>
- Read long documentsstructuralconfig: 262,144 context
Facts
20 fields
Listed in Decision models · LLMs
Facts
- Intern-Decision-4B has 4.54B parameters. (safetensors metadata, observed 2026-10-24)
- Context length is 262,144 tokens (256K). (config.json max_position_embeddings, observed 2026-10-24)
- License: Apache 2.0 (friction: none). Permissive; patent grant. (license file, observed 2026-10-24)
- Created 2026-09-26, last modified 2026-10-22. (HF API, observed 2026-10-24)
| Pipeline tag | image-text-to-text |
|---|---|
| Library | transformers |
| Architectures | Qwen3_5ForConditionalGeneration |
| Model type | qwen3_5 |
| Dtype | bfloat16 |
| Params by dtype | BF16: 4,539,261,696 · F32: 3,840 |
| Size band | M |
| MoE | no |
| Context length | 262,144 (config.json max_position_embeddings) |
| Papers | arXiv:2610.22131 |
| License | Apache 2.0 (apache-2.0) |
| Hub snapshot | 2026-10-25 |
Servability
vLLM · SGLang
Servability
vLLM: supported · Qwen3_5ForConditionalGeneration · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · Qwen3_5ForConditionalGeneration · 773cb938fe9484a8f7e7e46fb1b1fe96412c59ed
vLLM registry (7050ef881ff0835f6890a47db1859a24c65481d6) · SGLang registry (773cb938fe9484a8f7e7e46fb1b1fe96412c59ed) · Matched by architecture class name against each project's registry on main; 'supported' means the architecture is registered, not that this checkpoint was tested.
Hardware fit
3 estimates
Hardware fit ESTIMATE
Weights-only memory for Intern-Decision-4B: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 11 GB · Single GPU (<24 GB) |
|---|---|
| FP8 (1 byte/param) | 5.4 GB · Single GPU (<24 GB) |
| INT4 (0.5 bytes/param) | 2.7 GB · Edge (<4 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.
Decision facts
catalog
Decision-model catalog curated
| How it was made | full fine-tune (whole weights published) |
|---|---|
| Lineage bucket | F - fine-tune of an open model (public/synthetic data) |
| Lineage evidence | board file inspection (adapter/head/weight files) + HF base_model field |
| Base model | Qwen/Qwen3.5-4B-Base |
| DI 0.2.1 score | 37.81 (independent (DI 0.2.1 board)) |
| ECE (board) | 0.0278 |
| Median latency | 44.2 ms (DI board, 1x RTX PRO 6000, single process) |
| Hosted price (input) | self-hosted/M |
| Source | https://huggingface.co/internlm/Intern-Decision-4B |
All variants
7 variants
Lineage
Base model: Qwen/Qwen3.5-4B
quantized: 3finetune: 3adapter: 1
Top variants (7 shown, ranked by downloads)
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| bombdefuser-124/Intern-Decision-4B-GGUF | quantized | GGUF | — | — | apache-2.0 | bombdefuser-124 | 0 | 8k |
| prithivMLmods/Intern-Decision-4B-GGUF | quantized | GGUF | — | — | apache-2.0 | prithivMLmods | 1 | 1.1k |
| mertkayacs/Wahler-4B | finetune | — | 4.54B | M | apache-2.0 | mertkayacs | 2 | 323 |
| yeaay/Intern-Decision-4B-GGUF | quantized | GGUF | — | — | apache-2.0 | yeaay | 0 | 418 |
| mertkayacs/Karar-4B | finetune | — | 4.54B | M | apache-2.0 | mertkayacs | 2 | 283 |
| mertkayacs/Deem-4B | finetune | — | 4.54B | M | apache-2.0 | mertkayacs | 1 | 277 |
| renlu/intern-decision-4b-tool-gate-v0.14.0 | adapter | — | — | — | apache-2.0 | renlu | 2 | 20 |
Repository changelog
5 commits
Repository changelog 5 commits
Papers
1 paper
Papers 1
arXiv links found in the model card and its tags. "Peer-reviewed" is shown only with evidence: a published venue in OpenAlex (measured) or the authors' own journal reference on arXiv (reported).
Similar models
3 models
Similar models
- Qwen3.5-4B · 4.66B same size band (M), same task (image-text-to-text), more providers, permissive license
- gemma-3-4b-it · 4.30B same size band (M), same task (image-text-to-text), more providers
- Qwen3-VL-4B-Instruct · 4.44B same size band (M), same task (image-text-to-text), more providers, permissive license
Sources
3 sources
Sources
Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:43 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:46 UTC · Summary: 2026-10-07 15:17 UTC · Last verified 2026-10-25
Follow Intern-Decision-4B
Watch this model
- /feeds/models/internlm--Intern-Decision-4B.xml(RSS 2.0) · every change to Intern-Decision-4B
- /feeds/families/internlm--Intern.xml(RSS 2.0) · every change across the Intern family
Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.