Article
CommunityJev Models & Reimplementations
SubQ Snap
Subquadratic's public-preview decision model for agent workflows, served on a System One-shaped POST /v1/systemone endpoint at $0.05 per million input tokens; the blog prints 61.7 on Decision Index 0.2.1 against a Jev cell of 57.9.
- Format
- Article
- Reading time
- 3 min
- Published by
- Community
- Author
- Jev News
- Last verified
- Oct 2026
- Platforms
- API
- Capabilities
- Requires API key
Highlights
- Reads up to 384K tokens of text and images; up to four images, 4 MiB each decoded and 8 MiB total, with a 13 MiB request body.
- Company speed line: 80 ms with almost no context, 250 ms at 32k, 660 ms at 100k, and 2.5 seconds at 384k.
- Blog table prints Snap ahead on 13 of 16 enterprise-mapped Decision Index tasks, with PhishNChips the largest gain at 47.1 against 25.1.
- Public preview, first come first served: 100 million free input tokens to the first 10,000 sign-ups, then possibly a waitlist.
- Desk tests got noul 0.98 on a Stripe ticket in 119 input tokens and a three-question call at confidence 0.90 in 235 tokens.
Watch out
The Decision Index figures are Subquadratic's own prints on edition 0.2.1; the desk notes the space is now v0.3.2 with no Snap row, so 61.7 remains a company claim. It is a hosted public preview with no licence or weights stated, latency may vary while the platform is tuned, and its LCR scores cannot be compared with published AA-LCR results.
Questions about this entry
Short answers drawn from the linked project or page.
A hosted decision model for agent workflows from Subquadratic, in public preview. It answers noul, choice, and score questions over text and images on POST https://api.subq.ai/v1/systemone, with the public-preview id subq-snap-preview.
The October 12 blog prints Snap at 61.70 on Decision Index edition 0.2.1 against 57.90 for Jev. The desk notes the space is now v0.3.2 with no Snap row, so the figure remains the company's own print.
Docs price input at $0.05 per million tokens with free output; one 119-token call costs about $0.00000595. The preview offers 100 million free input tokens to the first 10,000 sign-ups, and a request without credits returns 402.
The window is 384K tokens, printed as 393,216, and a request over or close to that limit fails with 400. The post calls that twelve times Jev, whose models page lists 32k for state plus the longest question and 64k for a request.
Two POSTs returned HTTP 200: urgency alone came back noul 0.98 in 119 input tokens at 1,195 ms, and a three-question call chose technical at confidence 0.90 in 235 tokens at 2,291 ms, with output tokens always 0.
On three of the 16 enterprise-mapped rows: ACOS by 0.6, API-Bank by 1.6, and When2Call by 7.6. The largest printed gain is PhishNChips at 47.1 against 25.1.
No. It uses the System One shape: state plus named noul, choice, or score questions. Choice supports 2 to 255 options and Score 2 to 10 ordered levels, and several questions share one request.
More Jev resources like this
A Chinese-language decision model that returns a probability for each fixed candidate option in one forward pass, with no generated text or chain of thought. It is a LoRA fine-tune of Qwen3.5-0.8B on 13,048,000 CJ-Dataset rows, benchmarked on CJ-Bench against Chinese-Jev.
Jev Models & Reimplementations#community#github#python
Community
yanqiangmiffy/chinese-decisionPython
yanqiangmiffy/chinese-decisionPythonA small distilled sibling of Perplexity's pplx-decider-v1.2: a Qwen3.5-4B decider that reads a state plus typed choice, noul and score questions in one pass, keeping most of the 27B model's Decision Index.
Jev Models & Reimplementations#community#open-models#multimodal
Community
Perplexity's 27B decision model rebuilt on Qwen3.8-27B with a joint request format that answers a whole request's questions in one sequence, reporting 68.09 on Decision Index 0.3 and 77.43 on vision.
Jev Models & Reimplementations#community#open-models#multimodal
Community
A 27B decision model from torchcast-ai that answers choice, noul and score questions together in one forward pass over the TypeSafe-compatible /v1/systemone format; the card reports Decision Index 0.2.1 of 65.00, +7.09 over Jev 1.13, at 33 ms per yes/no question.
Jev Models & Reimplementations#community#open-models#local
Community
The Ollama build of iapp's OpenThai-SystemOne: a 0.8B Thai and English decision model whose v0.3 weights are fine-tuned for Ollama's /v1/systemone letter readout, reporting a 74.3 macro across 13 public subsets for Q8_0.
Jev Models & Reimplementations#community#open-models#local
Community
A 27B decision model from StartLux Labs that takes text, JSON or images up to 262,144 tokens and returns a probability per option through the TypeSafe /v1/systemone format, reporting Decision Index 0.2.1 63.88 and 208 of 231 on JevBench public.
Jev Models & Reimplementations#community#open-models#local
Community
From the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.