Perplexity's open-source Decider v1.3 scores 70 on Decision Index at $68 per 1M tasks
ArtificialAnlys · x · 2026-11-14
Artificial Analysis reviewed Perplexity's newly released Decider v1.3 and Decider Flash v1.3 decision models:
- Decider v1.3 (27B): scores 70 on Decision Index v2.1, holding #6; ranks #4 on AA-Paperclip (69) and #3 on EvalSafe-ONET-AA (89).
- Decider Flash v1.3 (4B): scores 64, up to #10, a 4-point gain over v1.2 driven by 12-point improvements on both AA-Symmetry and When2Call.
- Both sit on the capability-vs-cost Pareto frontier at $68 and $34 per 1M tasks respectively.
Other details: 256K context, $0.02/$0.01 per 1M input tokens, text+image input, Apache 2.0 license, available via Perplexity's first-party API.
More from Models
- Grok Bot optimizations deliver ~7% more plan usage, up to ~15% on routine runs — mattyp · 2026-11-14
- Brain has 200T synapses yet forgets a penny; a 20T-param LLM forgets nothing — fleetwood___ · 2026-11-14
- Another Week of Waiting: Frontier Pace Still Controlled Despite Model Drops — patience_cave · 2026-11-14
- Google spotted testing temporary chat feature in AI Mode — gaganghotra_ · 2026-11-14
- Unsloth's 5-10% quant gain was measured at 256K context — re-check your budget — petrusenko_max · 2026-11-14
- OpenAI Said to Test New ChatGPT Interface Codenamed 'Nova' — btibor91 · 2026-11-14