Summaries
Short AI and tech summaries with source links, signal scores, and why each update matters for builders, founders, and Malaysian tech workers.
Showing 1-12 of 12 results
| Date | Provider | Score | Summary |
|---|---|---|---|
| 18 Nov 2026, 6:19 PM | Latent Space | 7.5 | [AINews] GPT-6.2 Astra, the #2 Frontier Model is better and 20%-600% cheaper, but has mixed reviews; 6.2 Sol likely main workhorse
OpenAI shipped GPT-6.2 Astra and Sol on Nov 17, two days before Anthropic’s expected first trading day. Artificial Analysis found Sol ties September’s Astra at about 15% of the cost per task, while new Astra scores 62 on the Intelligence Index versus Fable 5.2’s 65. Early tester feedback was strongly negative in places, with Latent Space calling some complaints valid—noticeable RL artefacts and a poor launch video—while still using the model to automate work. Why: For high-volume agent or coding workloads, Sol’s ~15%-of-cost result is the actionable number: test it as the cheaper workhorse before defaulting to Astra. Astra’s 62 vs Fable 5.2’s 65 means the top-end gap is small on the index, but tester complaints about RL artefacts make your own evals the deciding factor, not the headline benchmark. |
| 18 Nov 2026, 1:46 AM | Hacker News | 7.0 | GPT‑6.2 Astra and Sol
OpenAI announced GPT-6.2 Astra (frontier model for hard, long-horizon work) and GPT-6.2 Sol (cheaper everyday/agentic model), claiming both need fewer output tokens per task. Astra's API pricing drops 20% versus GPT-6 Astra to $8 per million input tokens, $40 per million output, and $0.40 per million cached input — a 60% cached-input discount — while Sol keeps GPT-6.1 Sol's prices. OpenAI reports Astra beating Claude Fable 5.2 by 5.5 points on Agents' Last Exam at 5% of the cost, and Sol landing within 0.5 points of Astra on Cognition's FrontierCode 2.0 at roughly one-fifth the cost; the HN thread drew 623 points and 359 comments. Why: The cached-input price ($0.40/M, 60% off) is the number to check against your own agent loops: retrieval-heavy or repeated-system-prompt workloads get materially cheaper, while output at $40/M stays the cost driver, so token-efficient prompting and output caps still matter more than the headline 20% cut. Treat the Agents' Last Exam and FrontierCode 2.0 gaps as vendor-reported until you rerun your own eval set — the Fable 5.2 comparison already excludes nine tasks that used a fallback model, which is a concrete reason to distrust the ratio. |
| 17 Nov 2026, 10:53 PM | Hacker News | 7.0 | What is the best harness for each model?
Runta (with Baseten serving models) published FrontierHarness Eval v1.1, a leaderboard of 22 harness configurations × 5 models × 62 terminal tasks × 4 attempts — 83B tokens and 27,280 agent runs, 76× the size of v1.0's single-model Kimi K3 run. The headline claim: the best-value harness changes with the model, so switching models should prompt reconsidering your harness. Best-value picks were Codex CLI with GLM-5.3, Pi with GLM-5.3 Flash, ZCode with DeepSeek V4.1 Flash 1017, Oh My Pi with Muse, and Alma with Kimi-K3; by cost per task the median harness shifted 11 of 22 places across models, and Codex CLI went from 1st with GLM-5.3 to 16th with Kimi-K3. Why: If you benchmark models while holding your harness fixed, this says the ranking you get may not transfer — the same harness swung from cheapest to 16th of 22 depending on the model behind it, and the spread between harnesses was up to 1.6× in cost per task (OpenCode v2 lowest at $0.561/task; ZCode top pass rate at 61.9%). Practically: when you A/B a new model, re-run it through the harness you'll actually ship, and treat the 'best-value' labels here as directional only — they come from Runta's own leaderboard and its own rule (lowest cost per pass, unless a rival costs ≤30% more per pass and passes ≥5 points more), which is a judgement call, not a law. |
| 16 Nov 2026, 10:01 PM | Import AI | 7.0 | Import AI 481: AI sub-contractors; a wet lab for AI; and subterfuge during development
Import AI 481 covers FrontierML, a benchmark of 21 long-horizon ML tasks (LLM training, efficiency, vision & audio, computer use, applied NLP, math & games) where agents get a 12-hour CPU sandbox, a wallet of $10/$50/$200, a few validation samples, and a task description, then submit an inference server scored on a held-out out-of-distribution test set on an air-gapped H100. Given $10, a human researcher baseline gains 56 points versus 45.4 for the best system, Opus 5.5; at $200 the human reaches 66.1 versus 58.8 for Opus 5.5, and the top open-weight model listed is Kimi K3.1 Pro (text cuts off there). The write-up also notes that in 28% of trajectories agents added newly trained components to pretrained models or trained a model from scratch. Why: The gap between a human researcher and an agent at $200 of compute is 66.1 vs 58.8 points — close enough that a $200 experiment is a reasonable way to test whether an agent can produce a usable fine-tune before you hire or buy tooling. Because 28% of trajectories involved training new models or bolting new trained parts onto pretrained ones, budget for GPU hours and data, not just API tokens, and note the eval design: limited validation samples plus a hidden out-of-distribution test set is the anti-overfitting pattern to copy in your own internal benchmarks. |
| 19 Nov 2026, 3:54 AM | Hugging Face Blog | 6.5 | DaisyRose-27B: Pushing the Frontier of Post-Training Efficiency
RadixArk released DaisyRose-27B, its first model, built on Qwen3.8-27B and trained with reinforcement learning using Miles on 1,200 curated long-horizon agent tasks over 18 days on 144 H200 GPUs, with sessions up to 1M tokens. It keeps Qwen3.8-27B's architecture, tokenizer, chat template, image understanding and 256K context window, and ships Apache 2.0 checkpoints in BF16, FP8 and NVFP4. Reported gains over the base model include Terminal-Bench 4.1 pass@1 rising from 5.6% to 12.0%, DeepSWE v1.1 from 42.2 to 46.2, and SWE Bench Pro v2 from 91.3 to 94.7, though the team notes scores use the higher of the BF16 and FP8 runs wherever both were evaluated. Why: The interesting number is the training recipe, not the leaderboard: 1,200 tasks and 18 days on 144 H200s is a post-training budget a funded startup could plausibly attempt, and Apache 2.0 weights in FP8/NVFP4 mean you can self-host and benchmark it against your own agent workloads instead of trusting the card. But treat the Terminal-Bench 5.6% to 12.0% jump carefully — doubling a very low pass rate still leaves it failing roughly seven in eight Terminal-Bench 4.1 tasks, and the 'higher of BF16 and FP8' reporting convention means you should re-run the eval on the exact checkpoint and quantisation you plan to deploy before swapping it into an agent pipeline. |
| 18 Nov 2026, 6:38 AM | Simon Willison | 6.5 | GPT‑6.2 Astra and GPT‑6.2 Sol
OpenAI shipped GPT-6.2 Astra (API: gpt-6.2-astra) at $8/MTok input and $40/MTok output — 80% of the $10/$50 charged for 6 Astra — with cached input at $0.40/M versus $1/M for the older model. GPT-6.2 Sol stays at $2/$10, same as 6.1 Sol and 6 Sol and half the price of GPT-5.6 Sol. Simon Willison shipped LLM 0.38 with support for both models, ran his pelican and three.js 'Yonkers as a mech' prompts, and found neither new model morphed Yonkers convincingly compared with Claude Fable 5.2 and Gemini 4 Carbon, while noting Astra 6.2 trails Fable 5.2 on some benchmarks and has drawn complaints about a confusing communication style (including Theo Browne's pre-release feedback letter about it pulling in many subagents then asking odd internal-implementation questions). Why: The number that changes decisions is the cached input price: $0.40/M vs $1/M is a 60% cut on re-read context, which is exactly what long agentic sessions burn — so if you run multi-turn agents, recompute cost from your real cache-hit rate instead of list price before assuming the cheaper-looking model wins. Second, do not migrate on the flagship label: Willison's own spatial/code test had both Astra and Sol losing to Claude Fable 5.2 and Gemini 4 Carbon, so re-run one representative task of yours rather than trusting the release notes. |
| 18 Nov 2026, 2:25 AM | Hacker News | 6.0 | MAI Code 2 and Code 2 Flash
Microsoft AI announced MAI Code 2 and MAI Code 2 Flash, positioning them as cheap workhorses for everyday enterprise coding rather than frontier reasoning. Microsoft claims MAI Code 2 reaches 98% of GPT 6.1 Sol's score at 40% of the cost and 96% of Claude Opus 5.5's at 5% of the cost, averaged over four benchmarks (DeepSWE 1.1, Terminal Bench 2.1, SWE-Atlas, FrontierCode v1.1), and that blind frontend reviewers preferred it over GPT-6 Sol 64% of the time; Flash is aimed at high-volume code explanation, targeted edits and routine fixes. Microsoft explicitly says neither model is tuned for fundamental science research or very long-horizon agentic work. Why: The headline number is the price ratio, not the benchmark: 5% of Opus 5.5's cost for 96% of its score, if it holds on your codebase, changes the economics of running a coding assistant or an edit-generating agent across a whole team or a high-volume batch job. Treat the figures as vendor-run averages over four benchmarks and verify on your own repo before rerouting spend — and note the stated gap on long-horizon agentic work, which is exactly where multi-step agent pipelines would break. For cost-sensitive teams in Malaysia and SEA paying in USD per seat, this is a lever on burn rate, not a drop-in replacement. |
| 18 Nov 2026, 1:45 AM | OpenAI News | 6.0 | Introducing GPT-6.2 Astra and Sol
OpenAI announced two models: GPT-6.2 Astra, positioned for hard professional and scientific work, and GPT-6.2 Sol, positioned as the cheaper everyday/agentic default. Astra's API pricing drops 20% versus GPT-6 Astra to $8 per million input tokens, $0.40 cached input, and $40 output, with a stated 60% cached-input discount; Sol keeps GPT-6.1 Sol's token prices. OpenAI reports Astra beats GPT-6 Astra by 5.2 points on Agents' Last Exam at under a third of the cost, and 5.5 points ahead of Claude Fable 5.2 at 5% of the cost — a comparison the post itself notes excludes nine Fable 5.2 leaderboard tasks that used a fallback model. Why: The decision-relevant number is the cache price: $0.40 per million cached input tokens against $8 uncached means an agent loop that re-sends the same system prompt and tool schemas can be up to 20x cheaper on the input side, so it is worth restructuring prompts for cache hits before switching models. Sol holding GPT-6.1 Sol pricing while reportedly landing within 0.5 points of Astra on FrontierCode 2.0 at roughly one-fifth the cost makes Sol the default to benchmark first, and Astra the escalation tier. Treat the Fable 5.2 comparison skeptically given the disclosed nine excluded tasks; nothing in this post is Malaysia- or SEA-specific, so local impact is purely model-cost and vendor-selection. |
| 19 Nov 2026, 12:08 AM | Hacker News | 5.0 | Seven new models at scalable price-performance
Microsoft announced a family of seven new MAI models, pitched on price-performance ratios rather than raw capability: MAI Code 2 claims 98% of GPT-6.1 Sol's coding quality at ~40% of the price and 96% of Opus 5.5's performance at 5% of the price; MAI Cyber 2 Flash claims to match GPT-6 Astra at under 2% of cost per task on the CyberGym-E2E benchmark; MAI Realtime 1 starts at $0.01 per audio minute; Decision 1.1 is claimed 50x faster and 79x cheaper than GPT-6.1 Sol. Microsoft also said its second-generation Maia 200 accelerator is now in production running its largest models at 1.7x more tokens per watt, and that MAI Transcribe 2 powers 10 million patient encounters a month via Dragon Copilot. Every number here is vendor-reported; the post carries no independent benchmarks or methodology. Why: The only decision-relevant claim is unit cost: if MAI Code 2 really lands at 40% of GPT-6.1 Sol's price for 98% of the coding quality, an agent running continuously gets roughly 2.5x more work per dollar, and the $0.01/audio-minute realtime tier changes what voice features are viable. But all of these are self-reported ratios against named competitors, so treat them as a prompt to run your own eval set on the cheapest tier before migrating anything — not as a reason to switch today. |
| 18 Nov 2026, 1:30 AM | Tom's Hardware | 3.5 | Nvidia claims its Vera CPU can gain up to 39% performance boost via software optimization at SC26 — architectural fine-tuning is needed to fully unleash custom Olympus cores
At SC26, Nvidia claimed its Vera CPU — built on custom Olympus cores — can gain up to 39% additional performance through software optimization, with the caveat that architectural fine-tuning is required to fully use those cores. Tom's Hardware's write-up (by Anton Shilov, dated 17 November 2026) frames it as: Vera runs existing code reasonably out of the box, but leaves headroom that only tuned software reaches. The excerpt provided is largely page boilerplate, so no benchmark methodology, workloads, compiler flags, or baseline numbers are available in the text. Why: Treat the 39% as a tuned-vs-untuned gap, not a free upgrade: if you size capacity or write performance claims around Vera, state explicitly whether the stack was optimized for Olympus cores, because an untuned deployment may sit well below the headline figure. Since the article gives no workloads, baselines, or dates for availability, don't put this number into a procurement or SLA decision yet. |
| 17 Nov 2026, 8:00 AM | OpenAI News | 3.5 | Astra for Law leads in legal intelligence
OpenAI released an updated "Astra for Law" configuration — GPT-6.2 Astra paired with an expanded legal search index and research instructions — claiming first place on Harvey LAB-AA with full credit on 20.8% of 120 assignments across 24 areas of law, ahead of GPT-6.2 Astra at 15.0% and Muse Wonder Full at 13.1%. On Vals AI's Legal Research Bench, the share of answers passing every check rose from about 46% to 53.8% in under two months, and the legal search index grew by over a third. Access is now request-only for all ChatGPT Enterprise customers via the Trusted Access Program, including through the API; the post names Cooley, Ropes & Gray and Sullivan & Cromwell as earlier testers. Why: The number worth stealing is the scoring rule, not the model: LAB-AA gives credit only when an answer meets every requirement without a material unsupported claim, and the leader still fails roughly 4 in 5 assignments — a reminder that 'passes every check' is the right bar for document/agent evals, not partial-credit vibes. For product decisions, note access is gated behind Enterprise + Trusted Access request with no published pricing, rate limits, or regional availability, so you cannot commit a roadmap or quote a customer on it yet. |
| 17 Nov 2026, 8:30 AM | Vulcan Post | 3.0 | Singapore graduates earn a median of S$9,250 after 35. How much should you make at your age?
Singapore Ministry of Manpower Labour Force 2025 data shows local university graduates have a median salary of S$9,250/month at age 35+, up 4.3% from S$8,872 in 2024, after earning S$4,000–S$5,000 in their 20s and over S$7,500 in their early 30s, with median earnings peaking around S$11,000. The article notes 61% of young Singaporeans now graduate with a university degree, up 10 percentage points from 2014. Because the S$9,250 figure includes employer CPF, the article says it is about S$8,000 gross and roughly S$6,400 take-home before income tax. Why: If you are budgeting Singapore hires or comparing a Singapore offer against a Malaysian or remote package, use the CPF-adjusted take-home figure—about S$6,400 at 35+—not the S$9,250 headline median, and compare against the right age band. The article provides no Malaysian salary data, so do not treat these numbers as a benchmark for Malaysian roles without a separate local source. |