← Explore

Posts tagged with benchmarks

Open Weight Weekly · ·4 min read

The 27B Model That Embarrassed 397 Billion Parameters

Alibaba shipped Qwen3.6-27B on April 22nd, and the benchmarks don't make sense.

qwen3.6dense-architecturedeltanet
Open Weight Weekly · ·5 min read

Hugging Face's ml-intern Pushed a 1.7B Model Past Claude Code on GPQA

Yesterday Hugging Face open-sourced a tool that should make every ML engineer either slightly nervous or very excited.

ml-internhugging-facesmolagents
Open Weight Weekly · ·4 min read

Blackwell Has 96GB of VRAM. vLLM and Ollama Disagree About What to Do With It.

NVIDIA's RTX PRO 6000 dropped a 96GB Blackwell card into the workstation market, and suddenly every open-weight model under 70B fits unquantized on a...

vllmollamablackwell
Neural Dispatch · ·5 min read

Meta's Muse Spark Ditches Open Weights and Bets Everything on Thought Compression

Meta just did the one thing nobody expected: it shipped a proprietary model.

metamuse-sparkthought-compression
Data Eng Daily · ·5 min read

The Part of Vector Search Nobody Benchmarks

Every vector database vendor publishes benchmarks showing sub-5ms latency on a million vectors. Unfiltered.

vector-databasemetadata-filteringbenchmarks
Neural Dispatch · ·3 min read

The Smartest Thing About NVIDIA's Quantum AI Models Isn't the Benchmarks

Quantum computing has a plumbing problem.

nvidiaisingquantum-computing
Neural Dispatch · ·4 min read

Claude Opus 4.7's Coding Gains Are Real. So Is Its Stealth Price Increase.

Anthropic dropped Claude Opus 4.7 yesterday, and the headline number is hard to ignore: 64.

anthropicclaude-opusswe-bench-pro
Agent Patterns · ·4 min read

The Leaderboard Is Hiding a 50x Price Tag

Your agent scored 82% on Terminal-Bench 2.0.

benchmarkscost-optimizationevaluation
Open Weight Weekly · ·5 min read

Llama 4 Scout Is the Most Downloaded Model of April. It's Also a Mess.

Llama 4 Scout hit 1.2 million downloads in its first two weeks on HuggingFace.

llama-4metamixture-of-experts
Neural Dispatch · ·5 min read

Stanford's AI Index 2026: SWE-bench Nearly Hit 100% and Entry-Level Dev Hiring Fell 20%

Stanford dropped its annual AI Index today, all 277 pages of it, and honestly it reads like three different reports that someone stapled together.

stanford-ai-indexbenchmarksswe-bench
Agent Patterns · ·5 min read

3.4 Billion Parameters and the Tool-Calling Paradox

Somebody tested thirteen local language models on tool calling last month and the winner was 3.4 gigabytes.

tool-callingbenchmarkslocal-llm
Neural Dispatch · ·4 min read

$14.3 Billion Later, Meta's AI Strategy Is Closed-Source and Fourth Place

Mark Zuckerberg spent three years convincing the developer world that Meta was the open-source AI company.

metamuse-sparkopen-source
Neural Dispatch · ·5 min read

MiniMax M2.7's Self-Evolution Is Genuinely Interesting. Its Open-Source Label Is Not.

MiniMax just dropped the weights for M2.

minimaxm2.7self-evolving
The Prompt Engineer · ·4 min read

The Benchmark Gap Closed. The Prompting Gap Didn't.

MiniMax M2.5 — an open-weight model you can run on your own hardware — hit 80.

benchmarksmodel-selectioncontext-engineering
GPU Economics · ·5 min read

AMD Just Cracked a Million Tokens Per Second

For the first time in the MLPerf inference benchmarks, AMD posted numbers that don't require mental gymnastics to interpret.

amdmi355xmlperf
Agent Patterns · ·5 min read

Tool Calling Doesn't Care About Your Parameter Count

A 3.4 GB model just posted a 97.

tool-callingsmall-modelsbenchmarks
Open Weight Weekly · ·4 min read

Gemma 4's Secret Weapon Isn't the 31B — It's the 26B That Acts Like a 4B

Google shipped Gemma 4 yesterday under Apache 2.

gemma-4mixture-of-expertsapache-2
The Prompt Engineer · ·5 min read

You Don't Have to Beg for JSON Anymore

I spent three months in 2024 building retry logic for a pipeline that extracted product data from GPT-4.

structured-outputconstrained-decodingjson-schema
Neural Dispatch · ·5 min read

Qwen 3.6-Plus Just Beat Claude on Terminal-Bench — But Read the Fine Print First

Alibaba dropped Qwen 3.6-Plus this week, and the headline number caught my attention: 61.

qwenalibabaagentic-coding
Data Eng Daily · ·5 min read

Stop Running Spark for 40 GB Jobs

Every quarter, someone on the team asks: "Do we really need this Spark cluster?" For most of the jobs running on it, the answer in 2026 is no.

duckdbapache-sparkbenchmarks