InternLM2.5-7B-Chat
Shanghai AI Lab · released 2024-07-01 · Other (free academic; commercial on application) license
InternLM2.5-7B-Chat is a July-2024 7B chat model from Shanghai AI Lab with strong math (vendor MATH 60.1) and 1M-context support; its GPQA (38.4) is modest and it predates the 2025 reasoning wave.
Key specs
| Type | Local open-weight |
|---|---|
| Parameters | 7.74B total |
| Architecture | internlm2 |
| Context window | 33K tokens |
| Knowledge cutoff | — |
| Modalities | text |
| Recommended backends | — |
| Minimum viable rig | — |
Benchmark scores
| GPQA Diamond | 38.4% |
|---|---|
| SWE-bench Verified | — |
| AIME | — |
| MMLU-Pro | — |
| BFCL v3 (tool use) | — |
| Composite score | 3.84 |
| Community rating | No reviews yet |
VRAM & disk per quantization
| Quant | VRAM | Disk | RAM | Context |
|---|---|---|---|---|
| Q4_K_M | 6 GB | 4.5 GB | — | 33K |
Strengths & weaknesses
Strengths: Strong math reasoning for 7B (vendor MATH 60.1); 1M-token long-context support; Tool-use / agent capabilities
Weaknesses: Modest hard-science QA (GPQA 38.4); Pre-2025 model, outdated vs current 7-9B peers