buildConfirmed2 publishers Musubi released PolicyLM-1.7B under Apache-2.0 to score messages against a written policy in a reported 22 milliseconds on an H100. Teams can rewrite rules without retraining, though every speed and accuracy figure so far comes from Musubi's own tests.
Perspective Coverage
3 publishers
- Builder
- Builder 38%
- Operator
- Operator 37%
- Investor
- Investor 25%
Reality
- Evidence45
- Adoption10
- Hype gap+25
- Incentives70
- Confidence60
buildOne report1 publisher Mistral's Vibe keeps its approval classifier on a Mistral model even when GLM 5.3 writes the code
Mistral has added Z.ai's GLM 5.3 to Vibe, its terminal coding agent, and moved the smart-approve classifier onto a fast Mistral model whatever model is active. Teams that pick GLM 5.3 to write code still get a Mistral model deciding when a person must approve an action.
Reality
- Evidence45
- Adoption
- Not scored
- Hype gap+5
- Incentives50
- Confidence45
buildOne report1 publisher Cantina's open-weights exploit model ran a 60-task security eval for $2.38
Cantina released apex-flash-1, an open-weights vulnerability-research model it says solved 40 of 60 tasks for $2.38, against $74.68 for Claude Opus 5 High. There is no hosted endpoint, so teams download the 321-billion-parameter weights, pay for their own inference and verify the numbers themselves.
Reality
- Evidence45
- Adoption
- Not scored
- Hype gap+35
- Incentives70
- Confidence40
buildConfirmed3 publishers Ai2 open-sourced AstaBrief 8B, which writes cited research reports in 51.1 seconds against 178.5 for Asta's Claude-powered mode. Labs that cannot send unpublished research questions to a hosted model can now run a cited-report generator on their own servers.
Perspective Coverage
3 publishers
- Builder
- Builder 55%
- Operator
- Operator 33%
- Investor
- Investor 12%
Reality
- Evidence55
- Adoption18
- Hype gap+22
- Incentives55
- Confidence60
buildOne report1 publisher Cantina publishes a security model built from 50 of its own vulnerability cases
Cantina released Apex Flash-1, an open-weights security model fine-tuned on 50 of its own vulnerability cases. Its only performance evidence is a 60-task evaluation the company ran itself, and a separate build is modified to refuse fewer requests.
Reality
- Evidence40
- Adoption
- Not scored
- Hype gap+15
- Incentives70
- Confidence40
buildConfirmed5 publishers Z.ai says every gain in GLM-5.3 came from post-training on an unchanged base. If that holds, refresh cadence for self-hosted weights is set by RL runs, not pretraining runs.
Perspective Coverage
5 publishers
- Builder
- Builder 58%
- Operator
- Operator 33%
- Investor
- Investor 9%
Reality
- Evidence40
- Adoption30
- Hype gap+35
- Incentives70
- Confidence55
Three US providers host Moonshot's Kimi K3 at a tenth the cost of going to the source
Kimi K3's 1.4 terabytes of weights take eight Nvidia GB300s just to sit in memory. Export controls keep those chips away from Moonshot. Modal, Fireworks and Baseten price the hosted result at $3 in and $15 out.
Reality
- Evidence32
- Adoption45
- Hype gap+33
- Incentives74
- Confidence40
Who eats the compliance bill: Sacks and Amodei argue thresholds while OpenAI pauses training
The public fight is billed as one about regulatory capture. The operative question is cost incidence, and one lab just showed what an unlegislated security bill looks like.
Reality
- Evidence36
- Adoption24
- Hype gap+32
- Incentives82
- Confidence41
GLM-5.3 Buys Buyers Time: Z.ai's Coding Model Cuts Tokens, Not the Closed-Model Lead
Z.ai says its 743B-parameter GLM-5.3 hits 34.5% on its own code bench using 22% fewer output tokens than GLM-5.2. The weights are still two weeks out.
Reality
- Evidence32
- Adoption18
- Hype gap+38
- Incentives76
- Confidence36