internlm/Intern-S2-397B model card

Fields

id

sources/hf-intern-s2-397b-card

URL

https://huggingface.co/internlm/Intern-S2-397B

Publisher

Shanghai AI Laboratory (InternLM)

Kind

repo

Published

2026-09-13

Accessed

2026-10-18

notes

Primary card for the Intern-S2-397B record. HF API createdAt 2026-09-13, 1,124 downloads at access. Shanghai AI Laboratory's scientific multimodal model, described as scaling pre-training, multi-task reinforcement learning and long-horizon agent RL. The comparison table is an image (figs/intern-s2-397b.jpg); the numbers recorded on the model were read from it by OCR on 2026-10-18 and are vendor claims. The figure's footnotes state TerminalBench 2.1 was evaluated with Terminus 2 and that SWE-bench-Pro used several anti-hack configurations to stop the model reading the answers. Columns in the figure are Intern-S2-397B, Qwen3.5-397B-A17B, DeepSeek-V4-pro, Kimi-K2.7-Code, GLM-5.2, GPT-5.5, Gemini-3.1-Pro and Claude-Opus-4.8. License Apache-2.0 per the repo.

Notes

Primary card for the Intern-S2-397B record. HF API createdAt 2026-09-13, 1,124 downloads at access. Shanghai AI Laboratory's scientific multimodal model, described as scaling pre-training, multi-task reinforcement learning and long-horizon agent RL. The comparison table is an image (figs/intern-s2-397b.jpg); the numbers recorded on the model were read from it by OCR on 2026-10-18 and are vendor claims. The figure's footnotes state TerminalBench 2.1 was evaluated with Terminus 2 and that SWE-bench-Pro used several anti-hack configurations to stop the model reading the answers. Columns in the figure are Intern-S2-397B, Qwen3.5-397B-A17B, DeepSeek-V4-pro, Kimi-K2.7-Code, GLM-5.2, GPT-5.5, Gemini-3.1-Pro and Claude-Opus-4.8. License Apache-2.0 per the repo.

Cross-references

This record references no other records.