Model comparison
cma-mini vs Claude 4 Sonnet
Capability radar, value bars, and side-by-side posture.
cma-mini
Not scored
value score
Claude 4 Sonnet
91
value score
cma-mini ctx
—
Claude 4 Sonnet ctx
200K
Capability radar
Value · context · multimodal · openness · speed posture
Value head-to-head
- Claude 4 Sonnet91
Attribute tape
| Field | cma-mini | Claude 4 Sonnet |
|---|---|---|
| Developer | User01110 | Anthropic |
| Context | — | 200K |
| Modalities | Text, Code, Image | |
| Openness | — | Proprietary Model |
| Speed | — | Fast |
| Price | — | $3/1M in · $15/1M out |
| Value score | — | 91 |
| Summary | cma-mini is a 1.01M-parameter experimental causal language model built around a Micro-Channel Attention architecture with a deterministic byte-level tokenizer and a 2,048-token context window. It is a research prototype intended to probe the limits of very small transformer-style models rather than serve as a production assistant or API. | Balanced Claude model for coding and agents. |