DeepSWEmini-swe-agent
67%±4%
- bench
- DeepSWE
- version
- v1.1
- split
- public 113 tasks
- harness
- mini-swe-agent
- effort
- max
- n
- 113
- metric
- Pass@1
- value
- 67%
- ci
- ±4%
- $/task
- $0.61
- tokens
- 73k out
- steps
- 102
gpt-5.6-lunacodex -m gpt-5.6-luna. See Codex models.Closed. Same 1.05M / 128k window as Sol and Terra. Effort ladder includes ultra on Codex; DeepSWE cites max.
Requests over 272k prompt tokens bill the whole request at 2x input / 1.5x output on OpenAI’s published surcharge. Continuum pricing.json records that; ccusage does not.
Use Luna for mechanical volume and cheap hosted GPT-5.6. DeepSWE official: 67% ±4% at max, $0.61 per task, 73k out, 102 steps. AA Intelligence Index 52. Vals hero index 59.88%.
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.
gpt-5.6-sol), Terra for balanced work (gpt-5.6-terra). Official DeepSWE row is max, not a silent default.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.
OpenAI does not publish Luna weights. No HF Files button. Get Plus is hosted inference.
Lab card and OpenRouter listing only. No Hugging Face Files button, because there is no official repo to point at.
Fetched OpenRouter catalog on 2026-08-19. hugging_face_id was empty.
Continuum hosts gpt-5.6-luna, plus Sol and Terra, on the 19 Aug 2026 public allowlist.
gpt-5.6-luna
click to select
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.
DeepSWE official: $0.61 per task at max (73k out, 102 steps). That is the cheapest official row among Opus / Sol / Fable / Luna.
Official OpenAI + OpenRouter + Continuum rate card, 19 Aug 2026: $0.20 / $1.20 per million. Cache read $0.02 /M. Cache write $0.25 /M. Web search $0.01 per call on the catalog. ≥272k surcharge is 2× input / 1.5× output. Sol’s official list is $5 / $30. Do not paste it here.
Artificial Analysis Intelligence Index 52, fetched 19 Aug 2026. The AA tile is that published integer. AA did not publish a DeepSWE-style CI or step count we will copy onto the chip.
| Claim | Source | As of |
|---|---|---|
| DeepSWE 67% ±4% at max | DeepSWE official board | 2026-08-13 |
| Official Luna $0.20 / $0.02 cached / $0.25 write / $1.20 out; ≥272k $0.40 / $0.04 / $0.50 / $1.80; Batch Supported | OpenAI pricing.md | 2026-08-19 |
| OR catalog $0.20 / $1.20, cache r/w $0.02 / $0.25, search $0.01, ≥272k $0.40 / $1.80 | OpenRouter catalog | 2026-08-19 |
| Hosted id gpt-5.6-luna | GET /v1/chat/hosted/models/public | 2026-08-19 |
| 5.8T week tokens (+19% WoW) | OpenRouter rankings This Week | 2026-08-19 |
| OpenRouter id openai/gpt-5.6-luna, context 1,050,000 | OpenRouter /api/v1/models | 2026-08-19 |
Same official DeepSWE harness (mini-swe-agent, public 113 tasks), fetched 2026-08-19. Different effort labels are the lab's own setting on that board, shown here rather than normalized.
Continue through OpenAI's model family with GPT-5.6 Sol.
Luna is the live week-volume slug. Sol stays a sibling card because Continuum hosts it and buyers still search that name.
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.
Verified against the live public allowlist on 2026-08-19. Base URL is https://continuumcode.ai/v1. Keys are cont_sk_ from Settings, Account, Inference API. Personal keys need Plus or above.
curl https://continuumcode.ai/v1/chat/completions \
-H "Authorization: Bearer $CONTINUUM_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"gpt-5.6-luna","messages":[{"role":"user","content":"Review this diff."}]}'