Model tracker / Decision models / Torchcast AI
Torchcast Decision 31B
Open-weight 31B decision model on Gemma 4 31B with a TypeSafe-compatible API; text only
Torchcast Decision 31B is an open-weight decision model from Torchcast AI (Apache 2.0, 31B, a full fine-tune of Gemma 4 31B (v1.0) parameters): it answers typed questions, such as yes or no, pick one or pick several, with probabilities instead of writing text. There's no hosted version: you run it on your own hardware. It scores 91.8 (self-hosted; seconds are the board's estimate; unranked until the board's 27 Oct draw is equated) on the independent JevBench Capability Score (board v1.7.46), where Jev, the reference, scores 77.1.
- Made by
- Torchcast AI
- Weights
- Open, Apache 2.0
- Size
- 31B, a full fine-tune of Gemma 4 31B (v1.0)
- Where
- Self-host only
- Images
- No, text only
- Context
- 16,384 (longer prompts keep the head and tail of the state)
Torchcast Decision 31B price
US dollars per million input tokens. Decision models don't write text, so none of the hosted ones charge for output.
| Where | Input | Output | Notes | Checked |
|---|---|---|---|---|
| Self-host only | Self-host | – | – | 2026-11-09 |
Torchcast Decision 31B benchmarks
| Benchmark | Result | What it measures | Reported by |
|---|---|---|---|
| JevBench v1.7.46 | 91.8 | Capability Score: accuracy and calibration on typed decisions, with Jev as the reference system | Independent |
| JevBench time | 0.34 s | Seconds per decision: measured end to end for hosted APIs, the board's estimate for self-hosted models | Independent |
| Reported speed | 48 ms median per public JevBench item on an H100, 74 ms on an RTX PRO 6000 | Latency on the maker's own hardware | Torchcast AI |
Makers' speeds are measured on their own hardware and aren't comparable with each other or with JevBench's timings.
Decision models like Torchcast Decision 31B
Every other decision model we track, by independent JevBench score.
| Model | Input / M | JevBench | Weights |
|---|---|---|---|
| Torchcast Decision 31B | Self-host | 91.8 | Open |
| Piccolo-27B AgentScope | Self-host | 86.0 | Open |
| Mammoth-1 Baseten (Base Labs) | $0.10 | 85.2 | Open |
| H2O-Lightning-31B H2O.ai | Self-host | 84.3 | Open |
| Drex Max v1.1 Nace.AI | $0.03 | 84.2 | Closed |
| Perplexity Decider V1.2 Perplexity | $0.02 | 80.8 | Open |
| d1-ultra Liquid AI | $0.042 | 80.5 | Closed |
| Kev-27B Jared Palmer | Self-host | 78.9 | Open |
| H2O-Lightning-4B H2O.ai | Self-host | 77.0 | Open |
| Piccolo-9B AgentScope | Self-host | 76.6 | Open |
| Clef Cloudflare | $0.24 | 75.3 | Open |
| Jev 1.14 TypeSafe AI | $0.042 | 75.0 | Closed |
| d1 Liquid AI | $0.02 | 74.4 | Closed |
| iota-decide-4b Macrocosmos | Self-host | 74.0 | Open |
| Mercury Decide Inception | $0.02 | 73.6 | Closed |
| Piccolo-4B AgentScope | Self-host | 73.6 | Open |
| OpenAI Decisions (GPT-6 Luna) OpenAI | $0.10 | 73.5 | Closed |
| Microsoft-Decision-1 Microsoft | $0.042 | 70.8 | Closed |
| Elephant-1 Baseten (Base Labs) | $0.035 | 70.2 | Open |
| Strands Decider 26B-A4B AWS Strands Labs | Self-host | 66.7 | Open |
| Clef-flash Cloudflare | $0.038 | 63.8 | Open |
| Perplexity Decider Flash V1.2 Perplexity | $0.01 | 61.7 | Open |
| Kev-9B Jared Palmer | Self-host | 61.5 | Open |
| Clef-omni Cloudflare | $0.15 | 58.3 | Open |
| Joey-1 Baseten (Base Labs) | $0.01 | 57.8 | Open |
| d1-3B Liquid AI | Self-host | 49.5 | Open |
| d1-omni-600M (experimental) Liquid AI | Self-host | 32.8 | Open |
| Laya Convai Innovations | $0.011 | 32.5 | Open |
| CLM-8B Stanford and NVIDIA | Self-host | 20.1 | Open |
| Clairic Cloudflare | $0.10 | – | Closed |
| d1.1 Liquid AI | $0.02 | – | Closed |
| d1-2B Liquid AI | Self-host | – | Open |
| d1-omni-2B Liquid AI | Self-host | – | Open |
| Drex 1.5 Nace.AI | $0.04 | – | Open |
| Drex 2.0 Nace.AI | $0.03 | – | Open |
| Kev-0.8B Jared Palmer | Self-host | – | Open |
| Kev-4B Jared Palmer | $0.042 | – | Open |
| Koa Decisions Salesforce | Self-host | – | Closed |
| Strands Decider 12B AWS Strands Labs | Self-host | – | Open |
| Strands Decider 2B AWS Strands Labs | Self-host | – | Open |
| Strands Decider E2B AWS Strands Labs | Self-host | – | Open |
| Strands Decider E4B AWS Strands Labs | Self-host | – | Open |
Compare them all in the decision models table →
Torchcast Decision 31B FAQ
How much does Torchcast Decision 31B cost?
Torchcast Decision 31B has no hosted, per-token price: you download it and run it yourself, so the cost is your own compute.
Is Torchcast Decision 31B open weight?
Yes. Torchcast AI publishes the weights under Apache 2.0, so you can run and fine-tune it yourself.
How fast is Torchcast Decision 31B?
Torchcast AI reports 48 ms median per public JevBench item on an H100, 74 ms on an RTX PRO 6000, measured on its own hardware. JevBench gives 0.34 seconds per decision, its estimate for self-hosting.
How does Torchcast Decision 31B compare with Jev?
On JevBench's Capability Score (board v1.7.46), Torchcast Decision 31B scores 91.8 against Jev's 77.1. Jev costs $0.042 per million input tokens; Torchcast Decision 31B is self-host only. JevBench is a one-person project that uses Jev as its reference, which favours cheap hosted models.