# Arete-27B vs Cola2 (slim)

> Arete-27B scores 55.1 (C) to Cola2's 33.9 (F) for open-weight models you run yourself. Prices, MCP, x402, uptime and agent notes side by side.

- Full: https://www.anchorterminal.com/compare/arete-27b-vs-cola2.md (~8,250 tokens) · this version ~630 tokens · JSON https://www.anchorterminal.com/compare/arete-27b-vs-cola2.json · canonical https://www.anchorterminal.com/compare/arete-27b-vs-cola2
- Index: https://www.anchorterminal.com/llms.txt · API: https://www.anchorterminal.com/api/v1/index.json · MCP: https://www.anchorterminal.com/mcp · Updated: 2026-11-10
- Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data)

Arete-27B scores 55.1 (C) on agent readiness against Cola2's 33.9 (F), and leads in every scored category. Both do open-weight models you run yourself.

- Arete-27B: C 55.1, rank #123 of 160 open-weight models · https://www.anchorterminal.com/tools/arete-27b.min.md
- Cola2: F 33.9, rank #160 of 160 open-weight models · https://www.anchorterminal.com/tools/cola2.min.md
- Arete-27B, good for Agents that want Qwen3.8-27B's coding and tool use without its handling of politically sensitive topics, on a 24 GB card or a 128 GB machine Ahead on Reach, 50 against 25; Risk, 55 against 40; Agent interface, 60 against 20; Documentation and upkeep, 63 against 58.
- Cola2, good for Researchers working on latent or diffusion language models who want a checkpoint to study

| Category | Arete-27B | Cola2 |
| --- | --- | --- |
| Capability (35%) | pending | pending |
| Reach (25%) | 50 | 25 |
| Risk (20%) | 55 | 40 |
| Agent interface (10%) | 60 | 20 |
| Documentation and upkeep (10%) | 63 | 58 |

| Fact (where they differ) | Arete-27B | Cola2 |
| --- | --- | --- |
| Vendor | Nous Research | ByteDance (Seed) |
| Licence | Apache-2.0 | None stated for the weights (the code is Apache 2.0) |

Why the edge:
- Agent interface, Arete-27B +40: Arete-27B 60 (Chat template 20, a Jinja file in the repository.); Cola2 20 (Chat template 20, A chat template in tokenizer_config.json (4,116 characters), which came with the Qwen3 tokenizer.)
- Reach, Arete-27B +25: Arete-27B 50 (Memory 10, the only build from Nous Research is the 54.7 GB BF16 release, which needs about 66 GB and fits the reference machine.); Cola2 25 (Memory 15, the only build is the F32 safetensors set, 35.07 GB, and 35.07 GB × 1.2 = 42.1 GB, within 64 GB.)
- Risk, Arete-27B +15: Arete-27B 55 (File safety 25, safetensors only and no remote code.); Cola2 40 (File safety 15, safetensors only, no .bin, .pt or .pth files (25).)
- Documentation and upkeep, Arete-27B +5: Arete-27B 63 (Model card 8, no limitations, context length, languages or cut-off.); Cola2 58 (Model card 8, Intended use and limitations (the out-of-scope list, and the unconverged latent model) and the architecture are on the card and the GitHub model c…)