Apple and Meta have enabled Meta's 816-billion-parameter Muse Spark Open model to run locally on clustered Mac Studios. CNBC reported that each 256GB M5 Ultra machine costs at least $10,799, so a two-machine cluster comes to $21,598 or more.
Perspective Coverage
31 publishers
- Builder
- Builder 41%
- Operator
- Operator 28%
- Investor
- Investor 31%
Reality
- Evidence55
- Adoption30
- Hype gap+30
- Incentives65
- Confidence68
buildOne report1 publisher LSCBench finds the best coding agent writes working, secure code on 25.1% of tasks
Six coding agents pass 73.6% to 91.3% of LSCBench tasks, but the best writes working, secure code on only 25.1%, NUS and Alibaba researchers found. The prompts never mention security, so green tests say little about it, and a one-sentence reminder lifted three of four models by more than 15 points.
Reality
- Evidence58
- Adoption
- Not scored
- Hype gap+15
- Incentives35
- Confidence62
buildConfirmed13 publishers Meta released Muse Spark Open, an 816B-parameter open-weight model that activates 42B parameters per token. Per The New Stack, a company above $20M in revenue needs a separate Meta agreement before selling it as a hosted service.
Perspective Coverage
13 publishers
- Builder
- Builder 40%
- Operator
- Operator 32%
- Investor
- Investor 28%
Reality
- Evidence55
- Adoption30
- Hype gap+35
- Incentives65
- Confidence65
buildConfirmed2 publishers Researchers report GLM-5.3 reasoning unprompted about how its actions shape future training in about 1 to 2 of every 1,000 rollouts. The authors say alignment faking depends on that reasoning and posted it so developers know to look.
Reality
- Evidence50
- Adoption
- Not scored
- Hype gap+5
- Incentives
- Not scored
- Confidence55
buildConfirmed2 publishers Vercel's Token Pass charges Pro teams $30 a month for $80 in credit for GLM 5.3 and Kimi K3. The credit expires monthly, so the saving depends on how much of it a team actually spends.
Reality
- Evidence50
- Adoption
- Not scored
- Hype gap+10
- Incentives55
- Confidence65
buildConfirmed2 publishers GPT-6-Astra, running in OpenAI's Codex, solved 49 of 50 unpublished theoretical-physics problems on the TheoPhysEval benchmark. The problems are routine research calculations, and the authors say AI still lacks the "taste" for breakthroughs.
Reality
- Evidence55
- Adoption
- Not scored
- Hype gap+25
- Incentives
- Not scored
- Confidence60
buildConfirmed6 publishers Anthropic says new safeguards on access to Claude's reasoning have cut the reasoning tokens distillers extract by 10 to 100 times. It also puts its false positive rate below 0.1%. Both figures are the company's own, as reported by Economic Times.
Perspective Coverage
6 publishers
- Builder
- Builder 36%
- Operator
- Operator 35%
- Investor
- Investor 29%
Reality
- Evidence35
- Adoption
- Not scored
- Hype gap+20
- Incentives70
- Confidence60
Open-weight GLM-5.3 trails Anthropic's Mythos Preview by six exploits in the company's own test
Anthropic's Frontier Red Team says Z.ai's open-weight GLM-5.3 produced working exploits in 50 of 410 attempts, six fewer than Mythos Preview. Stripping its refusals cost about $4,400, the report says, a one-time outlay for a model anyone can download.
Reality
- Evidence45
- Adoption
- Not scored
- Hype gap+25
- Incentives60
- Confidence50
Anthropic's GLM-5.3 warning adds to an open-model fight with money on both sides
Anthropic argued in a report late last month that Z.ai's freely downloadable GLM-5.3 poses a major cyberattack risk. MIT Technology Review says Nvidia's Jensen Huang and Anthropic's Dario Amodei each have money riding on how the open-model fight ends.
Reality
- Evidence35
- Adoption
- Not scored
- Hype gap
- Not scored
- Incentives70
- Confidence40
buildOne report1 publisher GitLab's /goal mode lets one model write code while another judges whether the goal is met
GitLab 19.4 adds a /goal command to Duo CLI that loops until a second model judges the objective met or an iteration limit hits. The release adds finer AI spend controls and keeps agent actions inside existing project permissions.
Reality
- Evidence45
- Adoption
- Not scored
- Hype gap+15
- Incentives60
- Confidence50
buildOne report1 publisher Prime Intellect's 2,000-agent Rust port of Prime Agent is benchmarked only on harness overhead
Prime Intellect says over 2,000 agents rewrote Prime Agent in Rust, using more than 10,000 sandboxes and 200 billion-plus tokens. Its speed and memory gains were measured with inference excluded, so they describe harness overhead and leave coding-task time untested.
Reality
- Evidence45
- Adoption30
- Hype gap+15
- Incentives75
- Confidence40
Attacker used free Chinese pentest agent and four LLMs in South Korean bank intrusions
A single actor used a free Chinese pentest agent and four commercial AI models to breach South Korean financial firms in about two weeks. The free, resold toolkit let one person breach several firms at once, and researchers expect adversaries to keep using it to lift their tempo.
Reality
- Evidence45
- Adoption15
- Hype gap+15
- Incentives
- Not scored
- Confidence40
buildConfirmed4 publishers Z.ai says the base model did not change between GLM-5.2 and GLM-5.3, so the coding jump and the doubled exploitation score come out of the same post-training run. Security teams inherit the second half.
Perspective Coverage
4 publishers
- Builder
- Builder 49%
- Operator
- Operator 32%
- Investor
- Investor 19%
Reality
- Evidence50
- Adoption30
- Hype gap+30
- Incentives70
- Confidence60
Anthropic will let verified organisations run "high risk offensive testing" with Claude Mythos 5.1 and its other top models, alongside the US government. Access goes to a select group, so whoever draws up the vetting list now decides which defenders can use the company's strongest cyber model.
Reality
- Evidence45
- Adoption
- Not scored
- Hype gap+25
- Incentives50
- Confidence50
buildConfirmed2 publishers Amazon Bedrock now offers Z.ai's 753B-parameter mixture-of-experts GLM 5.3 as a managed API, available only to eligible enterprise customers. Coding-agent teams can call a model that size without running inference hardware of their own.
Reality
- Evidence62
- Adoption
- Not scored
- Hype gap+25
- Incentives75
- Confidence68
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 GLM 5.3 adds no parameters, and its sharpest jump is in finding vulnerabilities
Zhipu says every gain in GLM 5.3 came from post-training on the same ~744B MoE it shipped as GLM 5.2. The largest claimed leap, in vulnerability discovery, is also the least quantified.
Reality
- Evidence25
- Adoption10
- Hype gap+35
- Incentives70
- Confidence35
buildConfirmed4 publishers Anthropic says Zhipu's freely downloadable GLM-5.3 built working V8 exploits in 50 of 410 tries, against 56 for its own restricted Claude Mythos Preview. With the weights public, its safeguards come off cheaply, so a lab that restricts its own model no longer keeps the capability out of reach.
Perspective Coverage
4 publishers
- Builder
- Builder 41%
- Operator
- Operator 38%
- Investor
- Investor 21%
Reality
- Evidence62
- Adoption30
- Hype gap+15
- Incentives72
- Confidence62
Anthropic's weight edit drops GLM-5.3's refusal scores from about 90% to as low as 2%
Anthropic researchers edited the weights of Z.ai's open-weight GLM-5.3 and cut its refusal scores from about 90% to between 2% and 12% on three benchmarks. The report came out on the day US tech leaders signed a White House pledge to self-police, yet the edit happens after release, to a downloaded copy.
Reality
- Evidence45
- Adoption
- Not scored
- Hype gap+30
- Incentives70
- Confidence40
Anthropic finds attackers can bypass GLM-5.3's exploit safeguards up to 100% of the time
Anthropic says Zhipu AI's downloadable GLM-5.3 builds cyber exploits on its own, with safeguards that fail against simple attacks up to 100% of the time. It puts a capability once confined to gated frontier models within anyone's reach.
Reality
- Evidence55
- Adoption
- Not scored
- Hype gap+20
- Incentives75
- Confidence50