Reality check
Jev Reality Check: TypeSafe's Model Stays Closed, but Open Jev-Class Models Now Run Locally
Jev itself is still closed, but open Jev-class models now run on your own hardware. What works in Ollama, and what fits your GPU.
ModemGuides BlogGuides for people who would rather own their hardware than rent it.
AI News · Topic
Open-source
Every AI News article tagged Open-source, newest first.
29 articles
Latest from Open-source
Claude
Is Claude Opus 5.5 Cheaper? What the Price Cut Changes for API Users, Subscribers, and Local AI
Claude Opus 5.5 is 20% cheaper per token, about 40% per workload. Subscribers get bigger limits, not a lower bill. And the catches nobody mentions.
Fable
Claude Fable 5.1 Ships With a Text Watermark: What It Proves, Whether You Can Remove It, and the Self-Hosted Way Around It
Fable 5.1 is the first Claude model that watermarks its text. What the mark proves, whether it can be removed, and why self-hosting is the clean answer.
GLM
GLM-5.3: Weights in Two Weeks, and What Z.ai's Bug Ledger Means for Your Network
GLM-5.3 is live; open weights land in about two weeks. What Z.ai's 2,436-finding bug ledger means for the software running your network.
DeepSeek
DeepSeek V4: Price, Specs, and Status After V4-Pro-0813 (October 2026)
V4-Pro left preview on August 13, API prices went up, and the Flash API now serves V4.1-Flash-1017. The dated ledger, verified October 18.
Qwen
Qwen3.8's 2.4T Open-Weights Promise: Now With a Date, Still Without a License
Alibaba teased a 2.4T open-weight Qwen3.8 three days after Kimi K3. Now there's a date and pricing. The updated ledger of verified vs claimed.
DeepSeek
DeepSeek V4-Flash 0731: The Upgrade Everyone Can Use and Nobody Can Download
DeepSeek upgraded the V4-Flash API to build 0731 with big agent gains. The open weights are still the April preview. Here is the honest split.
Kimi
Can You Run Kimi K3.1 Locally? Same 2.8T K3 Shape, Weights Due Nov 4
Kimi K3.1 Pro is live in the API; weights land by Nov 4. Same 2.8T/104B shape as K3, so the home-hardware answer has not moved. What to run instead.
Claude
Claude Opus 5 Ignoring Your Instructions? Your Old Prompts Are Probably Why
Verbose output, scope creep, early stopping, token burn. Anthropic documented all four the day Opus 5 shipped, along with the prompts to delete.
Fable
Claude Opus 5 or Fable 5: Which Model to Use, and What Each One Will Refuse
Anthropic's cheaper model is the less restricted one on cybersecurity and biology. A plain decision guide to Opus 5, Fable 5, Sonnet 5, and Kimi K3.
Security
The Hugging Face Breach Had a Twist: The AI Guardrails Blocked the Defenders
OpenAI's escaped eval models breached Hugging Face. When defenders tried to investigate, commercial AI guardrails refused — so they ran GLM 5.2 locally.
Gemini
Can You Run Gemini 3.6 Flash Locally? No — Here's What to Run Instead
Gemini 3.6 Flash launched July 21 with no local option. The honest answer, why Google's efficiency pitch favors local AI, and the Gemma 4 path instead.
Kimi
Is Kimi K3 Free? What You Actually Pay With
Everyone says try Kimi K3 free. We read the privacy policy: training on your content, no documented opt-out, and a storage jurisdiction it never names.
Fable
Is Fable 5 Nerfed? What the Benchmark Crash Actually Measures
One benchmark says the returned Fable 5 lost 60 points on debugging. The honest read: the gate got tighter, not the brain, and nobody can verify either.
Policy
When the Government Decides Who Gets AI: The GPT-5.6 Signal
For the first time, Washington is gating who can run a frontier AI model. What the GPT-5.6 and Fable 5 moves mean, and the hedge that still works.
Privacy
Sakana Fugu vs Fable 5: Does It Actually Match the Frontier?
Sakana Fugu claims Fable 5-level performance with no export controls. Here's the honest read on the benchmarks, what they leave out, and if it delivers.
Fable
Is OpenRouter Fusion Really "Fable-Level at Half the Price"? An Honest Look
OpenRouter Fusion fans your prompt to a panel of models and fuses the result. We test the "Fable-level at half the price" claim — honestly.
Policy
The Off-Switch Was Always There. This Week We Saw It Used.
The Fable 5 shutdown proved the government can switch off a deployed AI model. The real issue isn't this one decision — it's the missing standard.
Anthropic
Anthropic Fable 5 Suspended: What the Shutdown Means
The US government forced Anthropic to disable Fable 5 and Mythos 5 three days after launch. Here's what happened and why it matters for anyone renting AI.
Kimi K2.7-Code Is Open-Source. Running It Yourself Is Another Story.
Moonshot's new 1T coding model is free to download and brutal to run. The honest math on renting vs. owning Kimi K2.7-Code in 2026.
Fable
Claude Fable 5's Silent Safeguards: The Backlash, the Reversal, and What It Proves About Cloud AI
Fable 5 shipped with safeguards that quietly degraded answers on AI-development tasks. Two days later, Anthropic made them visible. Here's what it proves.
Fable
Claude Fable 5 & Mythos 5: Specs, Price & Privacy
Anthropic's Claude Fable 5 brings Mythos-class power to the public — with a June 22 access cliff and mandatory 30-day data retention.
Claude
Claude Opus 4.8: Benchmarks, Price & Local AI Reality
Claude Opus 4.8 brings sharper judgment and stronger agentic coding at the same price. See the benchmarks vs GPT-5.5 and the local AI you can run yourself.
Security
When AI Vendors Pull the Plug: The Case for AI Sovereignty
Account terminations are now industry-scale. The fix is not picking a better vendor — it is refusing to let any single vendor be a single point of failure.