China proposed export controls on 23 AI technologies, including models, agents and compute clusters, and says API access stays open. Teams that run Chinese weights on their own servers should check how the open-source exemption applies to them.
Perspective Coverage
9 publishers
- Builder
- Builder 23%
- Operator
- Operator 36%
- Investor
- Investor 41%
Reality
- Evidence66
- Adoption25
- Hype gap+20
- Incentives58
- Confidence64
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
buildOne report1 publisher GMI Cloud offers a free week of Alibaba's Qwen and Wan models to sell its inference API
GMI Cloud is giving developers seven days of free access to Alibaba's Qwen3.8-Max, Qwen3.8-Flash and Wan3.0 via an OpenAI-compatible API. For a company built on renting GPU capacity, the week tests whether developers will take its inference service as an API.
Reality
- Evidence50
- Adoption
- Not scored
- Hype gap+5
- Incentives70
- Confidence55
buildOne report1 publisher DeepSeek and Qwen3.8-Max score near a blind guesser on macOS bash 3.2's set -u cases
DeepSeek and Qwen3.8-Max scored 0.52 and 0.40 on bash 3.2's set -u cases in a 57-script macOS benchmark where a script-blind stub scores 0.446. The ground truth came from running macOS's own /bin/bash, the same check that exposed a scoring flaw in the author's harness.
Reality
- Evidence45
- Adoption
- Not scored
- Hype gap+20
- Incentives20
- Confidence40
Alibaba's QwenBook routes Windows, Android and Linux apps through a chatbot
Qwen's chat interface is how you open apps on the QwenBook, whether they belong to Windows, Android or Linux. The staff demonstrating it at Snapdragon Summit could not say how the AI reaches those environments.
Reality
- Evidence45
- Adoption6
- Hype gap+50
- Incentives60
- Confidence45
Xiaomi's MiMo-V2.6-Pro leads the open-weight index at $0.87 per million output tokens
Artificial Analysis scores the new Xiaomi model first among open weights and twelve points behind Claude Opus 5.5, and the cheaper Flash tier is the one an operator should put in front of a real queue.
Reality
- Evidence55
- Adoption30
- Hype gap+15
- Incentives72
- Confidence55
Twenty model calls turn a two-second step into a 45-second wait
Red Hat clocks the same 20-call agent task at roughly 45 seconds on a slow backend and about 13 on a fast one. Model choice for agents is turning into a per-call latency budget, with capability as one input.
Reality
- Evidence45
- Adoption
- Not scored
- Hype gap+10
- Incentives50
- Confidence55
buildOne report1 publisher Every Qwen3.8-Omni-Flash workflow ends in text your own tools have to execute
Alibaba's new omni model takes text, images, audio and video across a million-token context and answers in text plus function calls. Rendering and token budgeting stay in the stack you already run.
Reality
- Evidence56
- Adoption20
- Hype gap+26
- Incentives78
- Confidence55
buildOne report1 publisher Two open-weight launches clear frontier-tier numbers on vendor engineering claims
DeepSeek's V4.1 Flash claims a million-token context on an FP4 KV cache and Alibaba opened 2.4 trillion parameters. The seven reports collected around the two launches quote no latency, throughput or hardware.
Reality
- Evidence28
- Adoption12
- Hype gap+40
- Incentives72
- Confidence35
buildOne report1 publisher Nex-AGI's runnable N2.5 tiers ask for two H100s or sixteen H200s
Publishing weights and publishing something a team can actually deploy are different acts, and N2.5 lands on both sides of that line depending on which tier you pick and whether its weights exist yet.
Reality
- Evidence45
- Adoption15
- Hype gap+30
- Incentives68
- Confidence50
buildOne report1 publisher Commerce Agent Bench decides pass or fail by reading the mock services after the agent stops
The Accio team's verifier ignores what the agent says it did and inspects the containers it leaves behind. That design is the news; the two-pass open-weight lead Alibaba reports from it sits inside its own harness spread.
Reality
- Evidence44
- Adoption22
- Hype gap+30
- Incentives82
- Confidence52
buildOne report1 publisher Four-bit weights leave 6 GB on a 24 GB card for KV cache and vision tensors
Meta's Muse Glimmer 30B and Alibaba's Qwen3.8-27B both landed in August under pure Apache 2.0 and both fit one 24 GB GPU, so the deployment question moves off licence terms and onto how you spend the memory that is left.
Reality
- Evidence16
- Adoption12
- Hype gap+52
- Incentives62
- Confidence20
buildOne report1 publisher Pooling three passes turns DeepSeek Pro's 17 findings into 28 of 32
Aikido spent 11.7 billion tokens rediscovering 32 fresh CVEs with ten models, three attempts each. The number that should move a scanning budget is the marginal cost of the second and third pass.
Reality
- Evidence44
- Adoption18
- Hype gap+32
- Incentives76
- Confidence52
Frontier scores at $2/$6: Grok 4.6 ties GPT-5.6 on one evaluator's index for a fifth the output price
Artificial Analysis puts Grok 4.6 at 61 on its Intelligence Index, level with GPT-5.6 Sol, at $2/$6 per million tokens. The same pages record 48 seconds to first token.
Reality
- Evidence56
- Adoption24
- Hype gap+24
- Incentives63
- Confidence50
Meta's coding agent has two prices: pay 18x more, or let it train on your repository
Muse Spark 1.2 costs roughly 18 times less if Meta may train on your prompts. The deepest cut, 75x, sits on cached input, which is where an agent holding your codebase in context spends.
Publishers:deeplearning.ai
Reality
- Evidence56
- Adoption16
- Hype gap+22
- Incentives78
- Confidence48
A 27B laptop model scores like a rented one, and thinks three times as hard to do it
Alibaba's Apache-2.0 Qwen3.8-27B fits in about 17GB and matched near-frontier scores, per Artificial Analysis. It also burned 3.7x the median output tokens getting there.
Reality
- Evidence62
- Adoption64
- Hype gap+18
- Incentives60
- Confidence55
buildOne report1 publisher Stripe paid $7B for a router. Glean is selling the choice not to call a model at all
Glean says its customers mostly turn on automatic model selection to control spend, not to improve answers. The routing layer, not the model, is where enterprise AI budgets now get decided.
Reality
- Evidence32
- Adoption64
- Hype gap+34
- Incentives82
- Confidence44
buildOne report1 publisher Qwen3.8's 27B dense checkpoint is the one operators can actually host
Alibaba scheduled a 27-billion-parameter vision-language model for August 14, alongside an already-published 2.4-trillion-parameter MoE. The smaller file is the consequential one.
Reality
- Evidence30
- Adoption10
- Hype gap+35
- Incentives75
- Confidence32
buildOne report1 publisher The zero-false-accept memory result belonged to the proxy, not the model
A verify-on-read experiment rerun across 14 live models on a fingerprinted 50-fact set found false-accept rates up to 0.38, and run-to-run noise wide enough to swallow a prompt fix.
Reality
- Evidence57
- Adoption14
- Hype gap−12
- Incentives31
- Confidence44