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.
LatestIBM4.2-30B
IBMText / LLM

IBM's Granite 4.2 30B adds reasoning and tool calls

The new dense model brings optional step-by-step thinking and function calling under a permissive Apache-2.0 license.

Aug 7, 2026
NotableApache 2.0
Granite 4.2 30B

IBM has published Granite 4.2 30B, the latest addition to its Granite line of open language models. The release, available on Hugging Face, is a dense 30-billion-parameter model that ships with support for reasoning-style "thinking" and tool calling, and it carries the permissive Apache-2.0 license.

The combination is notable for teams building agentic systems. Optional reasoning modes let a model work through problems in intermediate steps before answering, while native tool calling makes it easier to wire the model into external functions, APIs, and retrieval pipelines without heavy prompt engineering.

Why it matters

IBM has positioned Granite as an enterprise-friendly, openly licensed alternative to closed frontier models, and the 30B size sits in a practical middle ground:

  • Large enough to handle structured reasoning and multi-step tasks
  • Small enough to be deployed on more modest hardware than the largest open models
  • Apache-2.0 licensing that permits commercial use and fine-tuning with minimal restrictions

For developers already invested in the Granite ecosystem, the arrival of reasoning and tool-calling capabilities in a dense model broadens the range of applications the family can serve. As always, real-world evaluation on your own workloads is the best gauge of whether it fits; the model card on Hugging Face is the authoritative source for configuration and usage details.

Sources

  • ibm-granite/granite-4.2-30b

    Hugging Face

    Visit
OlderMiniMax Opens Its Music3 Text-to-Music ModelMiniMax · Music · 2 months agoNewerQwen releases 2.4T-parameter open MoE with 95B activeQwen · Alibaba · Text / LLM · 2 months ago

Get the model

Hugging Face

Specs

Parameters30B
Context window131K tokens
Size58.6 GB
PrecisionBF16
ArchitectureGraniteForCausalLM
LicenseAPACHE-2.0
Downloads37.2K
Likes129

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

    61.7 GB
  • 8-bit

    48 GB GPU (RTX 6000) or 2×24 GB · 64 GB Mac

    34.2 GB
  • 4-bit

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

    20.7 GB
Your machine
GGUF builds
  • 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

Text / LLMReasoning
2 versions — view changelog

The Weekly Weights

Every open release that mattered, one email a week.

0 comments

No comments yet. Be the first to weigh in.

More from IBM

All IBM releases →
IBM/Speech → Text

TII extends the Falcon family into speech recognition

The Abu Dhabi lab behind the Falcon language models releases Falcon ASR, an open speech-to-text system.

Oct 7, 2026
IBM/Text / LLM

IBM's Granite 4.2 Adds Reasoning to Open LLM Line

The latest update to IBM's Apache 2.0 model family leans into structured reasoning while keeping its enterprise-friendly licensing.

Aug 25, 2026
IBM/Text / LLMRuns anywhere

Liquid AI's LFM2.5 230M targets phones and robots

A 230-million-parameter language model built to run locally on constrained hardware like the Raspberry Pi.

Jul 1, 2026

More in Text / LLM

All Text / LLM →
Mistral AI/Text / LLMDatacenter

Mistral Large 4 arrives as a trillion-param MoE

The new flagship is a sparse mixture-of-experts model with roughly 49B active parameters per token.

Oct 6, 2026
NVIDIA/Text / LLM

Falcon-Emirati tunes an LLM for local dialect

TII's Falcon family gets a variant built around Emirati Arabic, aiming at culture and nuance rather than generic Gulf Arabic.

Oct 6, 2026
OpenAI/Text / LLMMulti-GPU server

Reflection releases Beam, a 501B open-weight model

The startup's first frontier-scale model ships with downloadable weights and a mixture-of-experts design aimed at reasoning.

Oct 5, 2026