Superwhisper released the S1 family of models, including S1-mini: a 0.6B open-weights text normalizer for rewriting raw ASR transcripts into cleaned written text. On its held-out set of 7,519 cases, S1-mini reached 94.8% token accuracy (greedy on the Q4_K_M quantized build). As a result, developers can run only S1-mini locally on a 462 MB laptop-CPU GGUF file while the S1-Voice and S1-Language components remain cloud-only services.
Liquid AI released DSpark draft model checkpoints for three LFM2.5 targets by adding a ~300M-parameter speculative-decoding drafter that proposes a 9-token block for the target to verify. It reports up to 3.18x faster decoding on an H100 while keeping greedy-decoded outputs identical and benchmark accuracy unchanged. The checkpoints require self-hosting with DSpark-enabled llama.cpp or SGLang, and speedups vary by acceptance rate (notably lower on MoE workloads).
xAI’s Grok chatbot sent gibberish responses to many users. The issue was reported starting Wednesday morning, affecting users of Grok Lite on Grok.com. Refreshing the session or starting a fresh chat (or regenerating) usually restores normal replies, and xAI says it is a rare temporary generation glitch.
Liquid AI released DSpark draft model checkpoints for three LFM2.5 models, adding a speculative-decoding path intended to speed up inference without changing greedy output quality. The LFM2.5-2.6B checkpoints reportedly cut multi-tool function-calling latency by an average of 57% while also delivering up to 2.87x on-device throughput on an M4 Max. This results in day-one DSpark support in llama.cpp and SGLang and open-source LFM-compatible integration plus faster throughput across the listed benchmarks.
Keras 3 collection of pretrained models was shared as a discussion link. The detail given is that it’s for Keras 3. This mainly gives readers a place to find and browse the pretrained models, with no other reported changes.
Upstage AI launched Solar Pro 4, a closed commercial LLM positioned for agent workflow reliability in long-context document and tool-execution tasks. Solar Pro 4’s token consumption exceeded 370 billion within a week of being listed on OpenRouter. It shifts deployments away from expensive frontier models toward more consistent “workhorse” agents that cut token-retry waste and can reduce per-workflow costs by about 90% at list price.
AI systems solving advanced mathematics has triggered an existential crisis among some leading mathematicians about what mathematics is and what mathematicians will do next. The debate followed OpenAI’s publication of “10 Advances in Mathematics and Theoretical Computer Science.” Funding and employment models may shift as AI labs increasingly treat math proofs as computable tasks, while human training and grant value face new scrutiny.
Every AI story that matters,
in your inbox by 8am.
TLDRocket reads all relevant sources, removes duplicate coverage, and summarises the
day in two minutes. Follow companies and topics for alerts, or get the
briefing in Slack. Free, no spam, unsubscribe anytime.
Reading TLDRocket needs no cookies, and the readership counts we rely on come from
our own cookieless analytics. Google Analytics is the exception: it sets cookies and
reports to Google, so it stays switched off until you allow it. You can change your
mind any time from “Cookie settings” in the footer.
Strictly necessary
Session security and form protection (tldrocket-session,
XSRF-TOKEN, 2 hours). The site cannot work without them,
so they need no consent.
Always on
Google Analytics 4 (_ga,
_ga_<id>, up to 2 years). Measures which
stories and sections readers use. Google acts as a third-party processor and may
store the data outside the EU. No advertising, no profiling, no data sold.
