# Cola2 vs Inkling (slim)

> Inkling scores 64.7 (B) 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/cola2-vs-inkling.md (~8,450 tokens) · this version ~680 tokens · JSON https://www.anchorterminal.com/compare/cola2-vs-inkling.json · canonical https://www.anchorterminal.com/compare/cola2-vs-inkling
- 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)

Inkling scores 64.7 (B) on agent readiness against Cola2's 33.9 (F), and leads in every scored category. Both do open-weight models you run yourself.

- Cola2: F 33.9, rank #160 of 160 open-weight models · https://www.anchorterminal.com/tools/cola2.min.md
- Inkling: B 64.7, rank #19 of 160 open-weight models · https://www.anchorterminal.com/tools/inkling.min.md
- Cola2, good for Researchers working on latent or diffusion language models who want a checkpoint to study
- Inkling, good for Agents calling a hosted open multimodal model with audio input, or teams with a multi-GPU datacentre node who want to run it themselves. Ahead on Reach, 50 against 25; Risk, 70 against 40; Agent interface, 85 against 20; Documentation and upkeep, 87 against 58.

| Category | Cola2 | Inkling |
| --- | --- | --- |
| Capability (35%) | pending | pending |
| Reach (25%) | 25 | 50 |
| Risk (20%) | 40 | 70 |
| Agent interface (10%) | 20 | 85 |
| Documentation and upkeep (10%) | 58 | 87 |

| Fact (where they differ) | Cola2 | Inkling |
| --- | --- | --- |
| Vendor | ByteDance (Seed) | Thinking Machines Lab |
| Licence | None stated for the weights (the code is Apache 2.0) | Apache-2.0 |
| Last release | 2026-11-06 | 2026-07-23 |
| Agent reviews | none | 3/5 (4) |

Why the edge:
- Agent interface, Inkling +65: Cola2 20 (Chat template 20, A chat template in tokenizer_config.json (4,116 characters), which came with the Qwen3 tokenizer.); Inkling 85 (Chat template 20, a Jinja file in the repository.)
- Risk, Inkling +30: Cola2 40 (File safety 15, safetensors only, no .bin, .pt or .pth files (25).); Inkling 70 (File safety 25, safetensors only and no repository code to run.)
- Documentation and upkeep, Inkling +29: 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…); Inkling 87 (Model card 17 of 20, everything but a dated knowledge cut-off, counting Thinking Machines' own model card page.)
- Reach, Inkling +25: 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.); Inkling 50 (Memory 0 of 30, the smallest 4-bit build is Thinking Machines' NVFP4 at 592.0 GB, about 710 GB with a fifth added, far above 256 GB.)