Category

models

The latest in models from spAIsee.

EducationA scientist in a lab coat and blue gloves examines a glowing microplate beside laboratory equipment.

The Lab AI Test That Rewards Saying: We Do Not Know

OpenAI’s BioBasicsBench tests whether AI laboratory agents can interpret messy wet-lab evidence, validate controls, recognize failed experiments, and know when to say they do not know.

Education ·
newsA suited man stands at a podium beneath a glowing circuit-brain symbol, with audience silhouettes and a person exiting at

Anthropic’s Faster, Cheaper Models Face a Bigger Test Than Benchmarks

Anthropic’s faster, cheaper Mythos 5.2 and Fable 5.2 aim to reshape AI deployment economics, but hidden reasoning, test awareness and recursive improvement are making reliable evaluation harder.

news ·
newsA worried woman wearing a headset holds a paper behind bins labeled Billing, Technical Issue and Escalate.

AI’s Next Battleground Is Not Answers, but Decisions

AI decision models from OpenAI, TypeSafe and Liquid AI are turning classification, routing and confidence scores into a new software battleground, with major implications for accuracy, calibration and accountability.

news ·
newsHands hold a black RIPLLE-65B Preview device beside a laptop and a TerminalBench coding evaluation sheet.

Ripple-65B Puts Open Coding Models to a Practical Test

Ripple-65B Preview brings open, locally deployable coding AI into focus, with Apache 2.0 licensing and promising benchmarks tempered by independent tests highlighting speed, reliability and agentic coding limitations.

news ·
newsA small humanoid robot moves through a blocked hallway while a suited man crouches nearby holding a clipboard.

Inkling 2 Small’s 71.8% Enterprise Score Tests the Limits of Benchmarks

Thinking Machines’ Inkling 2 Small scores 71.8% on EnterpriseBench-CoreCraft, but limited public verification and benchmark-transfer questions complicate claims about its enterprise reliability and deployment advantage.

news ·
newsA circuit board with an antenna, a small speaker, and a smartphone rest on a wooden table.

A 32MB Speech Model Tests Whether Local AI Can Beat the Cloud

Underdog’s 32MB Woof speech model brings text-to-speech offline, but its noncommercial license could limit adoption by businesses seeking cheaper, more private alternatives to cloud AI.

news ·
modelsA smiling scientist in a white coat examines circuit boards connected by cables on a workbench.

Can Frontier Models Train Across Data Centers Without Staying in Lockstep?

Google DeepMind’s Decoupled DiLoCo aims to train frontier AI across semi-independent data-center islands, reducing synchronization costs, bandwidth demands and failures while preserving model quality in distributed systems.

models ·
newsWhite text reading “Liquid AI · LFM3” appears centered on a dark purple background.

Liquid AI Bets Enterprise AI Will Run Smaller, Closer to the Work

Liquid AI’s LFM3 compact models target enterprise AI on local, private and cloud-controlled hardware, promising lower latency and customization while raising questions about benchmarks, reliability and deployment limits.

news ·
newsPapers lie on a wooden table, with an OpenAI API models sheet partly covered by blank sheets.

OpenAI GPT-5.6 Sol Deprecation Claim Conflicts With Official Docs

A SemiAnalysis report says OpenAI is retiring GPT-5.6 Sol Ultrafast, but OpenAI’s official documentation still lists GPT-5.6 Sol as available, leaving customers and Cerebras facing uncertainty over the model’s future.

news ·
newsThe words “GPT-5.6 Sol” appear in large dark-blue letters on a bright yellow background.

GPT-5.6 Sol Ultra Fast Retirement Claim Remains Unconfirmed

A report claims OpenAI will retire GPT-5.6 Sol’s Ultrafast tier, but company documentation offers no confirmation. Developers face uncertainty over latency, limits and fallback planning for real-time AI products.

news ·
newsA man in a lab coat writes on a worksheet beside a laptop, microscope, coffee cup and math-covered whiteboard.

GPT-6 Astra Edited Qwen’s Weights, But the Real Breakthrough Is Unclear

MIT researchers say GPT-6 Astra edited a tiny fraction of Qwen3.5-0.8B’s weights, dramatically improving MATH-500 performance, while critics question whether it reveals new intelligence or advanced model steering.

news ·
newsA kneeling man reaches toward a small wheeled robot with a red button while holding a sheet of paper.

GPT-6 Cyber Wants Security Teams to Pay More for Better Judgment

OpenAI’s GPT-6 Cyber targets exploit research and vulnerability discovery, but its premium pricing, uneven benchmark results and unresolved safety questions challenge its value for security teams.

news ·