DeepSWE
No official DeepSWE mini-swe-agent row for this identity on the public board we fetched.
Canonical official weights are meta-llama/Llama-4-Maverick-17B-128E-Instruct. This card lives under /models/meta/. There is no /models/meta-llama/ URL.
HF API (fetched 19 Aug 2026): architecture Llama4ForConditionalGeneration, model_type llama4, gated manual, license_name llama4 (Llama 4 Community License, effective 5 Apr 2025). The raw config.json URL 401s without accepting the gate, so layer counts are not restated from a file we could not open. Hub base_model is meta-llama/Llama-4-Maverick-17B-128E.
Use Maverick when you want Meta’s open Llama 4 Instruct checkpoint and you can accept the Llama 4 Community License plus the Hub gate. AA Intelligence Index 14 is cited. We do not invent a DeepSWE percent.
No official DeepSWE mini-swe-agent row for this identity on the public board we fetched.
We did not open a vals.ai card for this identity. Chip omitted.
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 |
|---|---|---|---|---|---|
| OpenRouter list pair | $0.20 / $0.80 per 1M | Catalog row for meta-llama/llama-4-maverick. No DeepSWE $/task. | OpenRouter catalog | source | 2026-08-19 |
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. Official Instruct repo only above.
Accept the Llama 4 Community License on Hugging Face. This page does not host the weights.
Raw config.json and README.md still 401 without accepting the gate (fetched 19 Aug 2026). Public card HTML (200) states vendor copy “17 billion parameter” and “402B param”. Those are vendor claims on the public page, not Hub tensor counts. Hub safetensors.total is 401,583,781,376.
OpenRouter catalog 19 Aug 2026: meta-llama/llama-4-maverick, context 1,048,576, text+image→text, hugging_face_id matches this repo.
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-00055.safetensors | 21,474,836,664 |
model-00002-of-00055.safetensors | 10,737,418,416 |
model-00003-of-00055.safetensors | 4,946,722,608 |
model-00004-of-00055.safetensors | 21,474,836,664 |
model-00005-of-00055.safetensors | 10,737,418,416 |
model-00006-of-00055.safetensors | 21,474,836,664 |
model-00007-of-00055.safetensors | 10,737,418,416 |
model-00008-of-00055.safetensors | 21,474,836,664 |
model-00009-of-00055.safetensors | 10,737,418,416 |
model-00010-of-00055.safetensors | 21,474,836,664 |
model-00011-of-00055.safetensors | 10,737,418,416 |
model-00012-of-00055.safetensors | 21,474,836,664 |
Showing the first 12 of 55 safetensor shards. The tree also holds 14 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 43 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. First-party Hugging Face (gated) and OpenRouter meta-llama/llama-4-maverick.
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: $0.20 / $0.80 per million for meta-llama/llama-4-maverick. No DeepSWE $/task.
Artificial Analysis Intelligence Index 14, fetched 19 Aug 2026 from their Llama 4 Maverick page. The AA tile is that published integer. AA did not publish CI, steps, or $/task for this identity in a form we will copy onto the chip.
| Claim | Source | As of |
|---|---|---|
| HF safetensors.total 401,583,781,376; usedStorage 1,004,052,580,220; 55 shards; likes 505; created 2025-04-01T22:17:20Z; gated manual; license_name llama4 | Hugging Face API Llama-4-Maverick-17B-128E-Instruct | 2026-08-19 |
| Raw README.md and config.json still 401 without accepting the gate (retry 19 Aug 2026) | HF raw README/config Maverick Instruct | 2026-08-19 |
| Public card HTML (200): vendor “17 billion parameter”; Hub UI “Safetensors Model size 402B params”; Llama 4 Community License effective 5 Apr 2025. Not Hub safetensors.total. | HF public card HTML Maverick Instruct | 2026-08-19 |
| OR id meta-llama/llama-4-maverick, context 1,048,576, $0.20 / $0.80, text+image, hugging_face_id official Instruct repo | OpenRouter catalog | 2026-08-19 |
| AA Intelligence Index 14 | Artificial Analysis Llama 4 Maverick | 2026-08-19 |
| OpenRouter id meta-llama/llama-4-maverick, context 1,048,576 | OpenRouter /api/v1/models | 2026-08-19 |
| HF downloads 14,014 | Hugging Face API meta-llama/Llama-4-Maverick-17B-128E-Instruct | 2026-08-19 |
| 55 safetensor shards; shard bytes 803,167,703,320; tree files 69 | Hugging Face tree API meta-llama/Llama-4-Maverick-17B-128E-Instruct | 2026-08-19 |
| Spaces API returned 50 (capped at 50) | 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 Meta's model family with Llama 4 Scout.
OpenRouter splits Meta and meta-llama. This cluster collapses both onto the Meta hub. There is no /models/meta-llama/ URL.
The official DeepSWE board updated 13 Aug 2026 did not list Llama 4 Maverick. Omitting is the rule.
The Hub tagged arXiv:2204.05149. That id is CLIP. We cite the tag and do not invent a Llama 4 technical-report title.
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 meta-llama/llama-4-maverick.