AITOPAI热报
ModelsMulti-sourceSelected

AI21 Labs Post-Trains NVIDIA Nemotron 3.5 Super VL for Enterprise Clients

官方账号AI21 Labs@AI21Labs
Post· 11 sources
View the post
Why we picked it

The interesting signal is the delivery model: a lab embedding researchers with customers to fine-tune an open NVIDIA base model cheaply, instead of selling API access to a proprietary frontier model. Engineering and AI leads evaluating build-versus-buy for domain models can use the thread's results as a reference point. Deel's involvement gives it a hiring-and-HR use-case angle.

AI21 Labs' Forward-Deployed Researchers post-trained NVIDIA's Nemotron 3.5 Super VL into specialized models over a three-week engagement. They worked with three companies: Gong.io, Deel and Alice.io, one specialized model per use case. AI21 reports the base model was pushed to match or surpass frontier quality across 3 use cases, at a fraction of the cost of frontier approaches. Detailed results were published in a thread on X.

Original · AI21 Labs

1/5 Over the past three weeks, our Forward-Deployed Researchers post-trained @NVIDIAAI Nemotron 3.5 Super VL into specialized models together with @Gong_io, @deel and @alice_dot_io. Across 3 different use cases, we pushed the base model to match or surpass frontier quality, at a fraction of the cost. Results below 🧵

#AI21 Labs#NVIDIA Nemotron#fine-tuning#open-source models

Keep reading

Recent stories with the same tag or section
How-toBuildersSelected17:07
AI21 Labs releases two AI Engineer Fair talks online

The chunking talk signals that naive fixed-size splitting of documents is losing ground to structure-aware approaches in production RAG systems. Engineers building retrieval pipelines can use it to re-evaluate their chunking strategies, and anyone running vLLM in production gets a concrete debugging walkthrough. Watching both takes well under an hour.

#AI21 Labs#RAG#vLLM#prompt engineering
官方账号AI21 Labs@AI21LabsSource
ModelsUS vs ChinaMulti-sourceSelected10:07
Anthropic models lead completeness but trail OpenAI on medical benchmarks

Verbose medical answers earn completeness points but introduce more claims that can contradict the record, so length itself becomes an accuracy liability. Teams building clinical documentation or decision-support tools should weight length-adjusted scores over raw ones, since the top model changes depending on that choice. Against alternatives like GPT-6 Astra, Claude models suit use cases where covering every rubric point matters more than brevity.

#Claude Opus 5.5#Claude Sonnet 5.5#GPT-6 Astra#HealthBench
Event
官方账号Artificial Analysis@ArtificialAnlys11 sourcesSource
ModelsBuildersMulti-sourceSelected03:11
Recursive Releases Flint, an Open-Weight 40B Coding Model

An Apache 2.0 license means teams can run, fine-tune and ship Flint commercially without restrictions, a rare setup among coding models. Its 40B MoE design targets local deployment, so developers with a 64GB-class GPU can test it offline and benchmark the 220 tokens-per-second claim themselves. Anyone comparing local coding assistants against proprietary cloud options has a new open baseline to try.

#Recursive#Flint#open-source models#coding assistants
Event
TestingCatalog@testingcatalog7 sourcesSource
ModelsBuildersMulti-sourceSelected00:47
vLLM adds Day-0 support for Recursive's Flint model

A sparse 2.5B-active architecture hitting 500 TPS on one GB200 shows how far inference efficiency now goes without speculative decoding, a benchmark point worth tracking against dense models of similar quality. Teams serving long-context workloads can run Flint in vLLM immediately in either NVFP4 or BF16. Builders evaluating open-weight model serving stacks should compare its throughput on their own GPUs.

#vLLM#Flint#NVFP4#open-source models
Event
官方账号vLLM@vllm_project8 sourcesSource
ModelsBuildersMulti-sourceSelected00:13
Scale AI Post-Trains NVIDIA Nemotron 3.5 Super VL

Post-training a smaller open NVIDIA model to match frontier accuracy at 1/40 the inference cost is a concrete signal that domain-specific fine-tuning can undercut flagship APIs for production agent workloads. Teams running high-volume ticket or support agents should benchmark Nemotron 3.5 Super VL against their current frontier model on cost and latency. The Skydance collaboration shows enterprises bringing their own production data to post-training partners.

#NVIDIA#Nemotron#Scale AI#fine-tuning
Event
官方账号Scale AI@scale_AI11 sourcesSource
ModelsUS vs ChinaMulti-sourceSelected23:55
Claude Sonnet 5.5 and GPT-6 Astra tie on MLCR-AA leaderboard

The split matters more than the tie: Anthropic models dominate completeness, with Claude Sonnet 5.5 at 64.8% versus 42.2% for the best non-Anthropic model, while OpenAI models dominate accuracy, led by GPT-6.1 Sol at 96.7%. Teams deploying models for medical record review should pick based on which failure mode they can tolerate. Even top models still fail more than half of the benchmark's questions, so human review remains necessary.

#Claude Sonnet 5.5#GPT-6 Astra#Wisedocs#Qwen3.8
Event
官方账号Artificial Analysis@ArtificialAnlys11 sourcesSource