LLM News

Every LLM release, update, and milestone.

0
benchmarkNVIDIA

NVIDIA Vera Rubin NVL72 posts fastest time to train on six of seven MLPerf Training v6.1 benchmarks

In its first MLPerf Training appearance, NVIDIA's Vera Rubin NVL72 delivered the fastest time to train on six of seven benchmarks, with up to 2x the per-GPU performance of an optimized GB300 NVL72 baseline, according to NVIDIA. On DeepSeek-V3 671B it trained in 4.14 minutes on 1,024 GPUs versus 8.03 minutes for GB300 NVL72.

3 min readvia developer.nvidia.com ↗
0
product updateCohere

Cohere adds Video Intelligence to Compass for timestamped clip retrieval, recruits design partners

On November 16, 2026, Cohere introduced Video Intelligence in Compass and began recruiting design partners. It retrieves timestamped clips, transcripts and visual descriptions alongside documents, and Cohere reports a 4.4-point average lead over the strongest alternative in its video-ranking evaluation.

1 min readvia unite.ai ↗
0
analysisAnthropic

Anthropic says Claude Mythos Preview onboarded AMD MI355X GPUs in about three days, versus six to 12 months

Anthropic told IPO investors that Claude Mythos Preview did the software work to run its AI on AMD MI355X GPUs in about three days, a job it says historically took six to 12 months. The three days cover prompt processing for one model only, and no performance comparison with Nvidia has been given.

3 min readvia tomshardware.com ↗
0
product updateOpenAI

OpenAI launches healthcare computer-use agents in ChatGPT and API, reports 99% of 325 reviewed tasks rated clinically sa

OpenAI announced computer-use agents for healthcare on November 16, 2026, available through ChatGPT for Healthcare, ChatGPT for Clinicians and the OpenAI API. The company reported that 99% of 325 reviewed GPT-6 Astra tasks in electronic health record inboxes were rated clinically safe.

1 min readvia unite.ai ↗
0
benchmarkSakana AI

Sakana AI releases JFinBench, a 321-question Japanese finance benchmark from 2026 certification exams

Sakana AI has released JFinBench, a Japanese financial benchmark of 321 questions drawn from the 2026 editions of six certification exams. Across 12 evaluated LLMs, Kimi K3.1 Pro falls about 10 points behind GPT-6 Astra, and Sakana AI says failures stem mainly from missing knowledge of Japan's systems rather than calculation errors.

3 min readvia sakana.ai ↗
0
model releaseTencent

Tencent launches Hy4 flash at $0.07 per 1M input tokens, one-twentieth the per-task cost of Hy4

Tencent launched Hy4 flash on Monday, a lower-cost model for everyday work such as research, slide creation and spreadsheet editing. On Tencent's internal WorkIndex benchmark, it cost one-twentieth as much per task as the full Hy4 model. API list prices are $0.07 per million input tokens and $0.281 per million output tokens.

1 min readvia technode.com ↗