GPT-5.6 Luna and Qwen3.6 35B A3B carry source-labelled ratings with evidence and any approved estimates in their category breakdowns. Models may use different benchmarks and test settings. This is an indicative composite, not a controlled head-to-head comparison or community Elo. Admin-approved sentiment estimates fill categories without accepted benchmark results. Estimates are labelled and do not increase benchmark coverage.
5 direct measurements, category ratings, capability breakdown, and verified sources
No overlapping benchmark runs recorded yet for this metric.
Different class (2s+ vs 500ms–2s)
Same speed class (50–100 tok/s)
$1 / 1M tokens
CompareLLM category Elo ratings across every applicable domain list at default effort.
Ratings represent CompareLLM category rankings derived from disclosed evidence recipes.
Methodology & Recipe →Simulate monthly API production cost in USD (US Dollar).
Qwen3.6 35B A3B saves approx. $6.50/mo ($78/yr).
Full Technical Dossiers & Specs
Need architecture, licensing, context windows, or provider rate cards?
No comments posted on this matchup yet. Be the first to share an evaluation note!
Published results are normalized using declared scales and weighted by category benchmark family. The breakdown lists the actual inputs, sources and weights. Models may use different benchmarks and test settings. This is an indicative composite, not a controlled head-to-head comparison or community Elo.
Context separates them most, favouring GPT-5.6 Luna. Partial fit: GPT-5.6 Luna is scored without coding, reasoning; Qwen3.6 35B A3B is scored without coding, reasoning.
Raw benchmark → percentile among comparable active models → workload weight → weighted contribution.
Raw: No canonical data
Raw: No canonical data
Raw: Out $/1M tok $1.2/1M tok
Raw: Speed (tok/s) 55 tok/s · TTFT 2,215 ms
Raw: Context (tokens) 1.1M tokens
Raw: No canonical data
Raw: No canonical data
Raw: Out $/1M tok $1/1M tok
Raw: Speed (tok/s) 57 tok/s · TTFT 819 ms
Raw: Context (tokens) 256k tokens
