LLM WATCH June 2026 baseline
Futuristic AI Dashboard Visualization
Captured 2026-06-04

Open weights move toward agentic practicality

Frontier open-weight models are now defined less by sheer scale and more by coding usefulness, multimodal support, long-context reliability, and deployment ergonomics. The gap to closed-source has closed to ~3-6 months.

Executive Summary

June 2026 snapshot of open-weight momentum, market structure, and deployment priorities.

The week’s biggest signal is a structural shift: the open-to-frontier gap has closed to ~3-6 months. DeepSeek V4 Pro (MIT, 1.6T/49B active) leads the open-weight frontier. MiniMax M3 is the first open-weight model to beat GPT-5.5 on SWE-Bench Pro. Apache 2.0 dominates licensing at ~38% of new releases.

Open-weight leaders are clustered around DeepSeek, Qwen, Kimi, GLM, MiniMax, and Xiaomi. Chinese labs dominate by volume — Qwen alone accounts for 399.4M downloads (18.5% of top-1k HF traffic).

Every frontier release in 2026 uses MoE architecture. Long-context evaluation is maturing beyond headline window sizes to actual degradation profiling.

Core signal this week

The open-weight field has closed the gap to ~3-6 months. It is a multi-polar market divided by coding, long-context, multimodal, and price-quality slices. Apache 2.0 dominates at ~38% of new releases. Chinese labs lead by volume.

Current phaseAgentic practicality

Releases Explorer

Browse the newest open-weight launches and the closed-source context releases that frame them.

Filters:
CompareModel & CreatorParams (Total/Active)ArchitectureLicenseKey Highlights / Milestones
ModelOrganizationRelease DateContext / Performance Notes

Composite Rankings

Cross-slice rankings from the baseline's open-weight, benchmark, and price-quality snapshots.

Artificial Analysis / Composite Open-Weight Snapshot

Open-weight frontier signals from recent composite summaries.

WhatLLM - Open Source Top Tier

Price-quality and speed-context slices for the strongest open source options.

Benchmark Leaders

The strongest single-metric and category-leading results cited in the baseline.

BenchmarkLeaderScore / ValueType

Licensing Matrix

Commercial usability and license-risk context for the current baseline.

License TypeRepresentative ModelsCommercial CompatibilityStatus

Use Cases Guide

Pick a deployment goal to see the best match and runner-up for this baseline.

RECOMMENDED TARGET

Best pick and runner-up from the current baseline

PRIMARY PICKRecommended Model
RUNNER UPAlternative Weight
Deployment Criteria
Deploy locally or via orchestrator API.Verified June 2026

What to Watch For

Signals most likely to move the field over the next 30 days.

Sources Cited

Primary sources used to build the June 2026 baseline snapshot.

Compare Weight Snapshots

Select of 2 models to run side-by-side verification

Weight Verification Compare

Metric / Metric Spec
Creator/Organization
Release Snapshot Date
Total Weight Parameters
Active Inference Parameters
Architecture Structure
License Commercial Safety
Core Benchmark & Performance Milestones