OpenGPTEvidence-based model decisionsStarting candidate 1
StableGrok 4.5
grok-4.5xAICost per successful task · #1
08 · TASK RANKING
Cost-efficient production
Reduce successful-task cost without choosing by price alone.
- Official version
- 3
- Public evidence
- 3
- Data cutoff:
- Why these candidates
- Use the published cost-per-success view to form a shortlist, then measure total retries, latency, and operational overhead.
- What remains uncertain
- Published prices and workloads change; provider token accounting and your success criteria may differ.
- Public evidence
- Cost per successful task
- Data cutoff
Starting candidate
3 / 3Three official versions to test first
Public data narrows the field. It does not replace a controlled test on your own work.
Starting candidate 2
StableGPT-5.6 Luna
gpt-5.6-lunaOpenAICost per successful task · #5
Starting candidate 3
StableGemini 3.5 Flash
gemini-3.5-flashGoogleCost per successful task · #9
Public evidenceThe models use different public evidence scales. Missing evidence is not treated as zero, and OpenGPT does not combine unlike sources into one score.⌄
Grok 4.5
- Source
- LiveBench ↗
- Source rank
- #1
- Source value
- $0.128
- Mapping reviewed
GPT-5.6 Luna
- Source
- LiveBench ↗
- Source rank
- #5
- Source value
- $0.202
- Mapping reviewed
Gemini 3.5 Flash
- Source
- LiveBench ↗
- Source rank
- #9
- Source value
- $0.249
- Mapping reviewed
A practical starting point, not a universal winnerSelect two or three candidates for a side-by-side public data comparison.
The models use different public evidence scales. Missing evidence is not treated as zero, and OpenGPT does not combine unlike sources into one score.
Published prices and workloads change; provider token accounting and your success criteria may differ.
The evaluation opens with this task template and the first two selected models. You can edit every task before collecting answers.