DeepSWE
No official DeepSWE mini-swe-agent row for this identity on the public board we fetched.
Open-weight Command A. License is CC-BY-NC-4.0. Week tokens on this hub belong to North Mini Code (238B, #42 on /models Top Weekly), not this card. Continuum’s free Cohere id is that other model. AA Intelligence Index 7; vals hero 43.41%. No official DeepSWE row. Catalog max out is 8,192.
Open weights on the official CohereLabs repo we snapshot. Max completion 8,192 on OpenRouter is narrow for long agent traces. Catalog created unix 1741894342 (13 Mar 2025). 256,000 in / 8,192 out; text→text.
HF API (fetched 19 Aug 2026): architecture Cohere2ForCausalLM, model_type cohere2, gated auto. The raw config.json URL 401s without accepting the gate, so layer counts are not restated from a file we could not open.
OpenRouter catalog hugging_face_id on 19 Aug 2026 is CohereForAI/c4ai-command-a-03-2025. This card’s snapshot key stays CohereLabs/c4ai-command-a-03-2025. Cite the catalog discrepancy; do not silently migrate the Files button.
Use Command A when you want Cohere’s open checkpoint and you can live with the non-commercial license and the 8k max out. AA Intelligence Index 7; vals hero 43.41%. We do not invent a DeepSWE percent.
No official DeepSWE mini-swe-agent row for this identity on the public board we fetched.
vals.ai Command A card, fetched 19 Aug 2026: hero Index 43.41%, printed $2.50/10.00, 2 min 28 s latency. Accuracy Rankings bars printed 0.0% placeholders: omitted. A Mar 2025 Updates paragraph prints LegalBench 78.7%, MGSM 86.8%, AIME 13.3%, GPQA Diamond 29.3%. Those named percents are Updates prose, not the hero chip.
Ranked view: Best coding models: the independent leaderboard puts this row and every other card that carries an independent coding score on one board, with the confidence intervals left visible.
Copied from AA or vals HTML we opened. 0.0% placeholder bars are omitted. Do not average these into the hero tiles, and do not invent CI, $/task, or steps for Artificial Analysis.
| Bench | Printed | Note | Source | URL | As of |
|---|---|---|---|---|---|
| AA list pair | $2.50 / $10.00 per 1M | Printed AA pricing sentence. Matches OpenRouter and vals. | Artificial Analysis Command A | source | 2026-08-19 |
| AA throughput | 54 tok/s | Printed “At 54 tokens per second, Command A is notably slow (89).” Not a coding score. | Artificial Analysis Command A | source | 2026-08-19 |
| AA blended rate | $3.25 per 1M | Printed “For a blended rate (7:2:1 cache hit/input/output ratio), this is $3.25 per 1M tokens.” Not DeepSWE $/task. | Artificial Analysis Command A | source | 2026-08-19 |
| vals Index extras | 2 min 28 s latency | Printed Latency on the vals hero. Hero Accuracy is the 43.41% chip. | vals.ai Command A | source | 2026-08-19 |
| vals Updates LegalBench | 78.7% | Mar 2025 Updates prose. Not a 0.0% bar and not the hero. | vals.ai Command A | source | 2026-08-19 |
| vals Updates MGSM | 86.8% | Mar 2025 Updates prose. | vals.ai Command A | source | 2026-08-19 |
| vals Updates AIME | 13.3% | Mar 2025 Updates prose. | vals.ai Command A | source | 2026-08-19 |
| vals Updates GPQA Diamond | 29.3% | Mar 2025 Updates prose. | vals.ai Command A | source | 2026-08-19 |
cohere/north-mini-code:free is not this model. Week tokens on this hub belong to that other SKU.Plus is $25/mo with $25 weekly hosted usage. The Mac app stays free with your own keys. Get Plus is not Download for Mac.
Community quants: Community quants are community.
Gated (auto). The Files button still points at the official repo; you accept Cohere’s gate on Hugging Face, not here.
Raw config.json and README.md still 401 without accepting the gate (fetched 19 Aug 2026). Public card HTML (200) states vendor copy “111 billion parameter” / “111B param”. Hub safetensors.total is 111,057,580,032. Do not write 111B as a Hub tensor count.
This is the model repo, not Download for Mac. Downloads and likes from the Hugging Face API on 2026-08-19.
usedStorage and the sum of .safetensors sizes can differ (LFS pointers, non-shard files, duplicate copies). Cited separately. Source: recursive tree API, 2026-08-19.
| Path | Bytes |
|---|---|
model-00001-of-00049.safetensors | 6,291,456,144 |
model-00002-of-00049.safetensors | 4,932,527,624 |
model-00003-of-00049.safetensors | 4,278,215,728 |
model-00004-of-00049.safetensors | 4,932,552,312 |
model-00005-of-00049.safetensors | 4,278,215,728 |
model-00006-of-00049.safetensors | 4,278,215,728 |
model-00007-of-00049.safetensors | 4,932,552,312 |
model-00008-of-00049.safetensors | 4,278,215,728 |
model-00009-of-00049.safetensors | 4,278,215,744 |
model-00010-of-00049.safetensors | 4,932,552,328 |
model-00011-of-00049.safetensors | 4,278,215,736 |
model-00012-of-00049.safetensors | 4,278,215,736 |
Showing the first 12 of 49 safetensor shards. The tree also holds 10 non-shard files, none of them listed here. The snapshot was capped at 12 shards when the tree API was read on 2026-08-19, so the remaining 37 are on the Files tab, not on this page. Totals above are the full tree API sum, not this slice. Byte sizes are Hub tree API size fields, never invented.
Not listed as a Continuum hosted id. Free lane is North Mini Code, a different identity.
We list first-party and OpenRouter, plus Continuum only when the live public allowlist named the id. This is not a 15-host routing table.
OpenRouter catalog 19 Aug 2026: $2.50 / $10.00 per million, 256,000 / 8,192, created 1741894342. No cache fields. No DeepSWE $/task.
Artificial Analysis Command A page (19 Aug 2026) prints the same $2.50 / $10.00 pair, 54 tokens per second, and a blended rate of $3.25 per 1M at a printed 7:2:1 cache-hit/input/output ratio. vals prints $2.50/10.00 and 2 min 28 s latency.
Artificial Analysis Intelligence Index 7, fetched 19 Aug 2026. Same page prints $2.50 / $10.00, 54 tok/s (“notably slow (89)”), and a $3.25 blended rate at a 7:2:1 cache-hit/input/output ratio. The AA tile is the published integer 7. AA did not print a DeepSWE-style CI or step count we will copy onto the chip.
| Claim | Source | As of |
|---|---|---|
| OR catalog $2.50 / $10.00, 256,000 / 8,192, created 1741894342, hugging_face_id CohereForAI/c4ai-command-a-03-2025 (card snapshot stays CohereLabs/…) | OpenRouter catalog | 2026-08-19 |
| HF safetensors.total 111,057,580,032; usedStorage 222,134,818,885; 49 shards; likes 394; created 2025-03-11T09:10:05Z; arXiv:2504.00698; gated auto; CC-BY-NC-4.0 | Hugging Face API CohereLabs/c4ai-command-a-03-2025 | 2026-08-19 |
| Raw README.md and config.json still 401 without accepting the gate (retry 19 Aug 2026) | HF raw README/config Command A | 2026-08-19 |
| Public card HTML (200): vendor “111 billion parameter” / Hub UI “111B params”; license cc-by-nc-4.0. Not a Hub tensor count. | HF public card HTML Command A | 2026-08-19 |
| AA Index 7; $2.50 / $10.00; 54 tok/s; blended $3.25 at 7:2:1 cache-hit/input/output | Artificial Analysis Command A | 2026-08-19 |
| vals hero 43.41%; $2.50/10.00; 2 min 28 s; weights dropdown “Open Weights & Proprietary”; Updates LegalBench 78.7% / MGSM 86.8% / AIME 13.3% / GPQA Diamond 29.3% | vals.ai Command A | 2026-08-19 |
| OpenRouter id cohere/command-a, context 256,000 | OpenRouter /api/v1/models | 2026-08-19 |
| HF downloads 1,964 | Hugging Face API CohereLabs/c4ai-command-a-03-2025 | 2026-08-19 |
| 49 safetensor shards; shard bytes 222,115,221,536; tree files 59 | Hugging Face tree API CohereLabs/c4ai-command-a-03-2025 | 2026-08-19 |
| Spaces API returned 11 | Hugging Face Spaces / collections API | 2026-08-19 |
No other card in this cluster shares a published DeepSWE official row we can put next to this one. We will not compare on vendor-blog numbers.
Continue through Cohere's model family with North Mini Code.
The official DeepSWE board updated 13 Aug 2026 did not list Command A. Omitting is the rule.
No. Continuum’s free Cohere id is North Mini Code. Command A is a different identity with an 8k max out and a CC-BY-NC license.
This card’s Hub snapshot is CohereLabs/c4ai-command-a-03-2025. OpenRouter’s catalog hugging_face_id on 19 Aug 2026 is CohereForAI/…. We cite both and do not silently move the Files button.
Lab posts pick a harness, an effort, a split, and sometimes a private eval set. DeepSWE publishes the official mini-swe-agent row with cost, tokens, and steps. Scale SWE-bench Pro only counts when the public shared-harness board has a row we can fetch. A higher lab number is usually a different test, not a better one.
No Continuum hosted id is verified for this model, so this page does not invent a curl target. Use the first-party API or the OpenRouter slug cohere/command-a.