Cheapest Decision Models per 1,000 Decisions
Cost per 1,000 decisions is the hosted input price at 200 input tokens per decision, the size of a typical state plus questions. Models below the median Decision Index score are left out, so the list is the cheapest of the competent ones; hosted models the Index has not scored yet are kept.
Ranked list (top 23)
| # | Model | Decision Index | ECE | p50 latency | Per 1k decisions | Params |
|---|---|---|---|---|---|---|
| 1 | SimpleJev Qwen3.6-35B-A3B (Featherless) Featherless AIOpen weights | — | — | — | Free | — |
| 2 | Span-01 Lite Respan | — | — | — | Free | — |
| 3 | Perplexity Decider v1.1 (27B) PerplexityOpen weights | 62.8 | 0.071 | 104 ms | $0.0040 | 28B |
| 4 | Liquid d1 (hosted) Liquid AI | — | — | — | $0.0040 | — |
| 5 | Mercury Decide Inception | — | — | — | $0.0040 | — |
| 6 | Span-01 Respan | — | — | — | $0.0040 | — |
| 7 | Cloudflare clef-flash CloudflareOpen weights | 47.6 | 0.025 | 53 ms | $0.0042 | 10B |
| 8 | Kai 1 Hanzo AI | — | — | — | $0.0042 | — |
| 9 | meraGPT Decider 1 meraGPT (Okyasoft) | — | — | — | $0.0060 | — |
| 10 | Vansa-3.4 Vansa | — | — | — | $0.0068 | — |
| 11 | Nace.AI Drex v1.5 Nace.AIOpen weights | 49.7 | 0.090 | 54 ms | $0.0080 | 9B |
| 12 | Celeris-1 Decision Celeris | — | — | — | $0.0080 | — |
| 13 | ScaleDown Classify-1 ScaleDown | — | — | — | $0.0080 | — |
| 14 | Jev TypeSafe AI | 60.1 | 0.074 | 524 ms | $0.0084 | — |
| 15 | Cloudflare clef CloudflareOpen weights | 53.1 | 0.040 | 102 ms | $0.0084 | 28B |
| 16 | Microsoft-Decision-1 Microsoft | — | — | — | $0.0084 | — |
| 17 | Sage 1.3.0 Levanto Labs | — | — | — | $0.010 | — |
| 18 | Solar Decide Upstage | — | — | — | $0.010 | — |
| 19 | Solar Decide Flash Upstage | — | — | — | $0.010 | — |
| 20 | Nace.AI Drex v2.0 (35B-A3B) Nace.AIOpen weights | 55.9 | 0.018 | 276 ms | $0.012 | 35B |
| 21 | GPT-6 Luna Decisions (OpenAI Decisions API) OpenAI | — | — | — | $0.020 | — |
| 22 | Fastino GLiDE Fastino | 60.2 | 0.040 | 98 ms | $0.030 | 28B |
| 23 | Cloudflare clef-omni (30B-A3B) CloudflareOpen weights | 48.3 | 0.040 | 191 ms | $0.030 | 32B |
Scores, calibration (ECE) and latency from the Jev Decision Index; cost per 1,000 decisions at 200 input tokens each. Updated daily.
More decision-model rankings
Frequently asked questions
What is the cheapest decision model?
SimpleJev Qwen3.6-35B-A3B (Featherless) is the cheapest hosted decision model in our data that is not below the median Decision Index score.
How is cost per 1,000 decisions calculated?
Input price per 1M tokens × 200 tokens × 1,000 decisions. Decision models return probabilities rather than text, so output tokens are negligible.
Is self-hosting an open decision model cheaper?
At high volume, usually: small open models run on a CPU, and the cost is the compute you already pay for. See the open-weights and CPU rankings.