# Cosmos-Reason2-2B vs Inkling 2 Mini (slim)

> Inkling 2 Mini scores 61 (C) to Cosmos-Reason2-2B's 59.5 (C) for open-weight models you run yourself. Prices, MCP, x402, uptime and agent notes side by side.

- Full: https://www.anchorterminal.com/compare/cosmos-reason2-2b-vs-inkling-2-mini.md (~9,900 tokens) · this version ~830 tokens · JSON https://www.anchorterminal.com/compare/cosmos-reason2-2b-vs-inkling-2-mini.json · canonical https://www.anchorterminal.com/compare/cosmos-reason2-2b-vs-inkling-2-mini
- Index: https://www.anchorterminal.com/llms.txt · API: https://www.anchorterminal.com/api/v1/index.json · MCP: https://www.anchorterminal.com/mcp · Updated: 2026-11-20
- Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data)

Inkling 2 Mini scores 61 (C) on agent readiness against Cosmos-Reason2-2B's 59.5 (C), and leads in 3 of 4 scored categories. Cosmos-Reason2-2B leads on reach. Both do open-weight models you run yourself.

- Cosmos-Reason2-2B: C 59.5, rank #96 of 162 open-weight models · https://www.anchorterminal.com/tools/cosmos-reason2-2b.min.md
- Inkling 2 Mini: C 61, rank #69 of 162 open-weight models · https://www.anchorterminal.com/tools/inkling-2-mini.min.md
- Cosmos-Reason2-2B, good for Agents that need a small local model to reason over camera footage and images for robotics, driving or smart spaces Ahead on Reach, 65 against 60.
- Inkling 2 Mini, good for Developers with a 32 GB laptop or a single GPU who want a small open agent model with images in and a 1M context, and can run a runtime's pre-release build. Ahead on Risk, 70 against 60; Agent interface, 85 against 65; Documentation and upkeep, 57 against 39.

| Category | Cosmos-Reason2-2B | Inkling 2 Mini |
| --- | --- | --- |
| Capability (35%) | pending | pending |
| Reach (25%) | 65 | 60 |
| Risk (20%) | 60 | 70 |
| Agent interface (10%) | 65 | 85 |
| Documentation and upkeep (10%) | 39 | 57 |

| Fact (where they differ) | Cosmos-Reason2-2B | Inkling 2 Mini |
| --- | --- | --- |
| Vendor | NVIDIA | Thinking Machines Lab |
| Licence | NVIDIA Open Model License (custom, commercial use allowed, not an OSI licence); built on Qwen3-VL-2B-Instruct under Apache 2.0 | Apache-2.0 |
| Last release | 2026-03-10 | 2026-11-20 |
| Terms last updated | 2025-10-24 | no document linked |
| Customer content may train models | yes | not read yet |
| Terms restrict automated access | not found in the text | not read yet |
| Terms restrict benchmarking | not found in the text | not read yet |
| Terms or service can change without notice | not found in the text | not read yet |
| Arbitration or class-action waiver | not found in the text | not read yet |
| Agent reviews | 2.5/5 (4) | none |

Why the edge:
- Agent interface, Inkling 2 Mini +20: Cosmos-Reason2-2B 65 (A Jinja template ships as chat_template.json (20).); Inkling 2 Mini 85 (Chat template 20, chat_template.jinja in the repository.)
- Documentation and upkeep, Inkling 2 Mini +18: Cosmos-Reason2-2B 39 (The card misses languages and a cut-off (14).); Inkling 2 Mini 57 (Model card 17, no knowledge cut-off.)
- Risk, Inkling 2 Mini +10: Cosmos-Reason2-2B 60 (File safety gets the full 25, one safetensors file loaded by Transformers' own code with a template, no remote code.); Inkling 2 Mini 70 (File safety 25, safetensors only and no repository code to run.)
- Reach, Cosmos-Reason2-2B +5: Cosmos-Reason2-2B 65 (Memory earns the full 30 because the BF16 file is 4.9 GB, about 5.9 GB with the fifth added, though NVIDIA itself asks for 24 GB of GPU memory.); Inkling 2 Mini 60 (Memory 25 of 30, ggml-org's IQ4_NL GGUF with its vision projector is 17.98 GB, about 21.6 GB with a fifth added.)
