The Open Weights
LatestModelsLeaderboardsCompanies
Subscribe
The Open Weights

The daily record of open-source AI. New model releases, leaderboards, and what's coming next — written for people who ship.

Refreshed every 12 hours

Discover

  • Latest releases
  • New today
  • Trending models

Browse

  • All models
  • Companies
  • Categories
  • Leaderboards

About

  • About
  • Editorial policy
  • Hardware estimates
  • RSS feed
  • llms.txt
  • Newsletter

© 2026 The Open Weights. An independent publication.

PrivacyTermsSMSAggregated by Claude · curated by humans.
LatestPerplexity Aiv1-27b
Perplexity AiVision-Language

Perplexity releases a 27B model for multimodal routing

The open-weight 'decider' model is designed to classify queries and route them inside Perplexity's stack.

Oct 1, 2026
UpdateApache 2.0
pplx-decider-v1-27b

Perplexity has quietly published pplx-decider-v1-27b on Hugging Face, a 27-billion-parameter vision-language model built for decision-making and classification rather than open-ended chat. As the name suggests, the model is a "decider" — the kind of component that sits upstream in a pipeline and routes incoming requests to the right downstream system.

The release is dense model (not a mixture-of-experts) in the 13B–34B class, and it handles multimodal inputs, meaning it can weigh both text and images when making routing decisions. That fits Perplexity's product, where a single query might need web search, a direct answer, or a specialized tool, and where getting that triage right has an outsized effect on latency and cost.

Why it matters

Routing and classification models rarely get headlines, but they are increasingly where production AI systems live or die. A few reasons this release is notable:

  • It is published under a permissive Apache 2.0 license, so others can study or adapt it.
  • It targets a practical, unglamorous job — classification and routing — that most labs keep internal.
  • A multimodal decider hints at how Perplexity is handling image-bearing queries in its stack.

Perplexity has not accompanied the weights with extensive documentation or benchmarks, so the model's exact capabilities and intended integration points remain thinly described on the model page. Still, the move continues a trend of search and assistant companies opening up the supporting infrastructure around their flagship systems, even when the flagship itself stays closed.

Sources

  • perplexity-ai/pplx-decider-v1-27b

    Hugging Face

    Visit
OlderCloudflare releases Clef open-weight decision modelsCloudflare · Text / LLM · 2 months agoNewerAleph Alpha releases Kolibri, a sovereign reasoning modelAleph Alpha · Reasoning · 2 months ago

Get the model

Hugging Face

Specs

Parameters27B
Context window262K tokens
Size52.2 GB
PrecisionBF16
ArchitectureQwen3_5Model
LicenseAPACHE-2.0
Downloads3.8K
Likes137

Can you run it?

Runs on a 24 GB GPU at 4-bit.

  • BF16 (as published)

    80 GB GPU (A100 / H100) · 128 GB Mac Studio

    55.3 GB
  • 8-bit

    32 GB GPU (RTX 5090) · 48 GB Mac

    30.8 GB
  • 4-bit

    24 GB GPU (RTX 3090 / 4090) · 32 GB Mac

    18.8 GB
Your machine
  • KV cache sized for 8,192 tokens of context; longer prompts need proportionally more.

Based on Hugging Face weights metadata. Assumes an 8K context and ~1 GB runtime overhead; actual needs vary. How we estimate


Modalities

Vision-Language

The Weekly Weights

Every open release that mattered, one email a week.

0 comments

No comments yet. Be the first to weigh in.

More in Vision-Language

All Vision-Language →
SenseNova-Loop-9B
SenseTime/Vision-LanguageRuns on a laptop

SenseTime Releases Looped 9B Multimodal Model

The company's new SenseNova-Loop-9B pairs vision-language capabilities with a recurrent looped-transformer design.

Nov 17, 2026
Hunyuan Hy4-flash
Tencent/Vision-LanguageMulti-GPU server

Tencent releases Hunyuan Hy4-flash multimodal MoE

The lightweight, mixture-of-experts model adds vision and text instruction-following to Tencent's open Hunyuan lineup.

Nov 17, 2026
Mirage
Cactus Compute/Vision-Language

Cactus Compute's Mirage brings vision tasks on-device

The new open vision-language model targets detection, segmentation, and visual Q&A without relying on the cloud.

Nov 5, 2026