Skip to main content
Van Data Team

Blog

Interactive research tool

AI Model Benchmarks

Compare current models using source-verified access, context, pricing, and evaluation evidence.

131 models23 providersVerified 2026-10-13
Open the benchmark hub
InsightOctober 14, 2026

Claude Haiku 5.5 vs Haiku 4.5: What Changes When You Upgrade

Claude Haiku 5.5 beats Haiku 4.5 on every published test and costs less per token. But it thinks by default, waits longer before its first word at some settings, refuses some requests Haiku 4.5 answered and rejects several old API settings. Here's what changes and how to upgrade safely.

AI Models & News
InsightOctober 14, 2026

Claude Haiku 5.5: Real Cost, Benchmarks and When to Use It

Anthropic's Claude Haiku 5.5 lists at $0.10 and $0.50 per million tokens, 90% below Haiku 4.5. But a new tokenizer, thinking that's on by default and a price step at 100,000 tokens decide what you really pay. Here's the math, the independent scores and how to test it.

AI Models & News
InsightOctober 13, 2026

Russian Military AI: Moscow's Claims vs. the Evidence

Russia's defence ministry opened a combat-footage service for training military AI on October 12, 2026, and a Russian analyst told Sputnik that Palantir and Anthropic had supplied their systems to Ukraine. We checked each claim against public records: what Russia announced, what is on the record about Palantir and Anthropic, how autonomous these systems are, and what the data race means for teams that build AI.

AI Models & News
InsightOctober 13, 2026

Muse Wonder: Meta's Frontier Model, Benchmarks and Price

Meta released Muse Wonder and Muse Wonder Full on October 12, 2026, and says they reach the AI frontier. Here's what the two models are, what Meta's own benchmarks claim, what independent tests found, what Muse Wonder costs next to rivals, why Meta needs it for its Muse agent, and how to test it fairly before you switch.

AI Models & News
InsightOctober 12, 2026

Firmus IPO Pulled: What It Means for AI Compute Buyers

Nvidia-backed Firmus pulled its roughly US$5 billion Australian IPO on October 9, 2026, after orders faded in three days. Here's what happened, why investors balked at a US$30.6 billion valuation, what it says about the AI trade, and the questions every team renting GPU capacity should now ask its provider.

Cloud Cost & Platform Ops
InsightOctober 12, 2026

Claude Startups Approval: What Happened After the Pause

Van Data Team was approved for Claude Startups, then re-reviewed during Anthropic's pause and confirmed a day later. Here's what the emails said, why Anthropic paused the free Claude Team year and $1,000 API credits, what an approval is really worth, how to claim it and what to do if your approval was pulled.

AI Models & News

Latest articles

Start with the newest

A short list of recent Van Data Team articles. Open the archive when you want to browse older topics.

Show the full archive

Need more than ideas?

Turn the reading list into a scoped delivery conversation.

If one of these articles maps directly to your current workflow pressure, the next useful step is usually a review of the system, constraints, and next build decision.

Explore servicesTalk through your workflow