InternLM open-sources Intern-Decision: 4B multimodal model beats Jev 1.13.0
Spico197 · reddit · 2026-09-27
The InternLM team released Intern-Decision, a multimodal decision model family in 0.8B/2B/4B sizes, open on Hugging Face and GitHub. The 4B model outperforms Jev 1.13.0 in their tests.
On a single RTX 4090, local deployment reaches 30 FPS decision calls, opening room for real-time applications. The team also built a calibration benchmark using classic probability problems (dice rolls, Monty Hall): Jev's decision model turns out to be poorly calibrated, while Intern-Decision's predicted distributions stay closer to the golden distribution.
Related event: Shanghai AI Lab Open-Sources Intern-Decision Multimodal Decision Models(4 posts)→
More from Models
- Will OpenAI's rumored $500 plan just reprice what $200 Pro used to offer? — VraserX · 2026-09-29
- User Tests Suggest Free-Tier Sonnet 5.5 Underperforms Paid Version on Claude — chaseleantj · 2026-09-29
- Three GPT models all drew a pelican 'riding' a bicycle by carrying it on its back — jade_jade_jade_jade · 2026-09-29
- LLMs increasingly filter out self-promotion in AI search, tests across Grok, Claude and ChatGPT show — lilyraynyc · 2026-09-29
- One env var change broke ticket tagging: GLM 5.3 no longer accepts thinking off — sweetcake_1530 · 2026-09-29
- H Company launches Holo4, a family of generalist computer-use models for desktop, web and Android — RemiCadene · 2026-09-29