Kev 4B
4B
- input
- $0.035
- output
- $0.00
- Context
- 16Ktokens
- Released
- 3w ago
- Size
- 9.4Bparameters
Cost/1M tokens
Baseten's Elephant 1 is an open-weights decision model that its Base Labs research team fine-tuned from Qwen3.5-9B. Announced as one of the world's top-ranked decision models and the leading 9B decision model on the Jev Decision Index, it returns a probability distribution over the possible answers instead of generating text, which Baseten says can give up to a 10x speedup over a reasoning model. It is built as a generalist for classification-style decisions such as support-ticket routing, spam and phishing detection, and routing requests between smaller and frontier LLMs.
- Decision model
- Open weights · 9.4B
- Accepts images
How it compares
Family tree
Models trained from one another, as their makers describe it.
Qwen3.5-9BQwen · 2026
Frequently asked questions
Running locally
Elephant 1 needs an estimated 7.1 GB of VRAM at Q4, 12 GB at Q8 and 22.1 GB at FP16. All figures are estimates from the 9.4B parameter count, weights plus about 15% and 0.5 GB for a short context.
Yes. A 24 GB GPU fits at FP16 (estimated 22.1 GB), a 16 GB or 12 GB GPU fits at Q8 (estimated 12 GB), and an 8 GB GPU fits at Q4 (estimated 7.1 GB).
Yes at Q4, which is an estimated 7.1 GB. Assuming macOS lets the GPU use about 75% of unified memory, the smallest common Mac memory size that fits Q4 is 16 GB.
API and pricing
The listed price is $0.035 per 1M input tokens and $0.00 per 1M output tokens, from 1 paid provider tracked.
A typical request with 10,000 input tokens and 1,000 output tokens costs $0.00035 at the cheapest provider, BaseTen.
Over the last day, 1 endpoint had uptime data and 1 was at 99% uptime or better. The best uptime was 100.0%, by BaseTen.
Privacy and data
BaseTen is the provider with zero data retention, being ZDR eligible or 0-day retention.
Yes. Elephant 1 has open weights, so you can self-host it for full control of your data.
Capabilities and specs
Elephant 1 has a context window of 16K tokens (16,384) and a maximum output of 15K tokens (14,745).
Elephant 1 accepts text and image inputs. Its listed input modalities are text and image, with no audio, video or PDF support listed.
Quality and benchmarks
Clef Flash by Cloudflare at $0.0285 per 1M tokens, Kev 4B by Jared Palmer at $0.0315 per 1M tokens, and Tev1 4B Experimental by Together AI at $0.0315 per 1M tokens. They were matched on model kind, blended price and release date.
Clef FlashElephant 1 vs Clef FlashKev 4BElephant 1 vs Kev 4BTev1 4B ExperimentalElephant 1 vs Tev1 4B Experimental
Versions and lineage
Elephant 1 is fine-tuned from Qwen3.5-9B.
Licensing and access
Elephant 1 is open weights.
The Elephant 1 weights can be downloaded from the Hugging Face repository baselabs/elephant-1.
Basics
Elephant 1 was made by Baseten and released on October 21, 2026.
Not fully. The weights are open, and the listed API price is $0.035 input and $0.00 output per 1M tokens.
Generated from Vertical Tabs data, updated November 11, 2026.
Discussion
No comments yet
Be the first to share your thoughts.