Benchmark Guide
Terminal-Bench 2.1 vs 4.0: what changed?
Why coding-agent scores change across benchmark versions, how harnesses affect results, and what the available 2.1 data can—and cannot—tell you.
Nov 2, 2026 · 8 min readIdeas behind better decisions.
Model analysis, benchmark guides and practical advice for builders.
How to evaluate coding agents on your own repository
Choose real tasks, control the harness, review patches, and measure the cost and reliability of accepted changes.
Read articleAll articles
31 articlesFrontier Analysis
Meta is back in the AI race
Muse Glimmer puts Apache 2.0 weights back inside Meta’s comeback. Spark is chasing the frontier; Glimmer is bringing agents home.
Aug 10, 2026 · 18 min readDefinitive Guide
Kimi K3: benchmarks, pricing, API, context window, and open-weight status
The definitive guide to Moonshot AI's 2.8T model: verified specifications, benchmark caveats, $3/$15 API pricing, 1M context, architecture, access, code examples, and who should use it.
Jul 20, 2026 · 24 min readFrontier Analysis
The open-model gap is not one gap
Open weights are roughly four months behind the aggregate frontier. Across coding, agents, reasoning, cost, privacy, and deployment, that average hides the decision that actually matters.
Jul 16, 2026 · 15 min readBenchmark Guide
Which coding LLM benchmark should you trust?
SWE-bench, Terminal-Bench, LiveCodeBench, and SciCode rank different jobs. A practical guide to choosing the benchmark that matches the code you need written.
Jul 16, 2026 · 14 min readCost Analysis
The cheapest coding model is not always the cheapest
Price per token ignores retries and review. Learn to calculate the cost of an accepted change without combining unrelated benchmark results.
Jul 16, 2026 · 12 min readAnalysis
Cost per Task Is the New Agentic AI Model Benchmark
Artificial Analysis Intelligence Index v4.1 changes the model-selection question from "which model is smartest?" to "what does a completed agent task cost?" A data-backed guide to Agentic Index, cost per task, cache pricing, Terminal-Bench 2.1, τ³-Bench Banking, and GDPval-AA v2.
Jun 16, 2026 · 13 min readMonthly Briefing
New AI Models May 2026: The Frontier Took a Breath, Architecture Took the Stage
After April broke the ceiling (GPT-5.5 at 60.24, Opus 4.7, DeepSeek V4, Kimi K2.6), May went quiet on scale and loud on architecture. SubQ shipped the first commercial subquadratic LLM with a 12M context. Zyphra dropped an 8B MoE trained on AMD. OpenAI made GPT-5.5 Instant the new ChatGPT default.
May 13, 2026 · 14 min readEssay
The world is built out of a few narrow places
A long letter on why I built AI Bottlenecks. EUV machines, indium phosphide wafers, CoWoS slots, gas turbines, and the four physical chokepoints gating the 2026 to 2030 AI buildout. The constraint is never at the loud end of the stack.
May 5, 2026 · 15 min readAnalysis
The state of AI at the end of 2025: reasoning won, agents arrived, and the race got tighter
A synthesis of Artificial Analysis's 2025 Year-End State of AI report. Reasoning models took the leaderboard, the cost of intelligence collapsed 100x, coding agents went mainstream, and the frontier got more contested — not less.
May 5, 2026 · 14 min readBreaking
DeepSeek V4 is here: the open model that made Jensen Huang's 'horrible outcome' real
DeepSeek V4-Pro and V4-Flash just dropped. 1.6T MoE, native 1M context, MIT weights, trained on Huawei Ascend. Priced at ~1/20th of Opus 4.7. The model Jensen warned about eight days ago on Dwarkesh — live on Hugging Face.
Apr 24, 2026 · 17 min readAnalysis
Kimi K2.6 is here: the open model that refuses to clock out
Moonshot AI just shipped Kimi K2.6. 1T parameters, 262K context, 4,000 tool calls in a single run, and benchmarks that put it shoulder to shoulder with GPT-5.4 and Claude Opus 4.6.
Apr 21, 2026 · 11 min readWeekly Briefing
New AI Models April 2026: Anthropic Won't Ship Its Best. Open Source Will.
Claude Mythos is locked behind a 50-company firewall. GLM-5.1 beat GPT-5.4 on coding under MIT license. Gemma 4 went Apache 2.0. The full breakdown of AI's first week of April.
Apr 8, 2026 · 18 min readMonthly Roundup
New LLMs March 2026: GPT-5.4 Tied for #1. Nobody Talked About It.
GPT-5.4 matched Gemini 3.1 Pro within 0.01 points. NVIDIA unveiled trillion-parameter infrastructure. Anthropic clashed with the Pentagon. Nine text models shipped, seven open-weight. Full breakdown.
Mar 24, 2026 · 14 min readAnalysis
Gemini 3.1 Pro Preview: what the .1 actually means
Google pushed a mid-cycle update to Gemini 3 Pro on February 11. We measured every benchmark delta: +4.2pp on SWE-Bench, +5.2pp on AIME 2025, #1 on LM Arena. Here is what changed and who should switch.
Feb 20, 2026 · 16 min readOpinion
The white-collar existential crisis: how AI killed the meaningless job
AI is not destroying jobs. It is exposing that most white-collar work was never truly meaningful. The K-shaped economy is splitting fast, agents are hiring humans, and the next 2-3 years decide which side you land on.
Feb 16, 2026 · 18 min readGuide
DeepSeek models: what to use, what to skip, and where to run them
A data-driven DeepSeek hub that connects benchmarks to provider reality: pricing, speed, and time to first token across hosts.
Jan 15, 2026 · 12 min readGuide
MiniMax models: what to use, what to skip, and where to run them
A MiniMax hub for builders. Understand M1 vs M2 vs M2.1 and compare providers by price, speed, and time to first token.
Jan 15, 2026 · 12 min readDeep Dive
The unspoken bottleneck reshaping artificial intelligence
The narrative around AI has long revolved around compute power. As we enter 2026, a quieter shift is underway. Memory, not chips, is becoming the constraint.
Jan 8, 2026 · 22 min readYear in Review
2025 AI Year in Review: The Year Intelligence Became Infrastructure
Twelve months ago, we were debating whether AI could reason. Now we're debating who owns the reasoning.
Dec 26, 2025 · 22 min readDeep Dive
The Open Source Revolution: How December 2025 Changed Everything
DeepSeek V3.2 hit 96% on AIME 2025. Xiaomi dropped a frontier model. GLM-4.7 claimed the coding crown.
Dec 26, 2025 · 18 min readIndustry
Three Forces That Broke OpenAI's Moat
OpenAI declared "Code Red" in December 2025. China's 15 open-weight models, the efficiency revolution, and custom silicon converged.
Dec 3, 2025 · 14 min readAnalysis
The State of LLMs: December 2025
Analysis of 114 models reveals a market in transition: benchmarks saturating, open-weight matching proprietary at 10x lower cost.
Dec 3, 2025 · 16 min readComparison
Gemini 3 Pro vs GPT 5.1 vs Claude Opus 4.5
The AI arms race explodes in November 2025 with three frontier releases in 12 days. Benchmarks, pricing, and where each dominates.
Nov 25, 2025 · 18 min readComparison
Kimi K2 Thinking vs ChatGPT 5.1: Reasoning Showdown
Moonshot AI's open-weight MoE takes on OpenAI's proprietary GPT-5.1. Architecture, benchmarks, pricing.
Nov 18, 2025 · 15 min readAnalysis
Kimi K2 Thinking: How Open Weights Are Catching GPT-5
Moonshot AI's K2 Thinking lands a 67 on Artificial Analysis, sets agentic records, proves open weights compete on reasoning.
Nov 2025 · 13 min readGuide
Top Open Source LLMs October 2025: Complete Guide
DeepSeek V3.1, Qwen3-235B, GLM-4.6 - performance benchmarks, pricing, and deployment insights.
Oct 2025 · 18 min readOpinion
Why AI Outputs Are Turning Into Repetitive Slop
Lessons from Andrej Karpathy: model collapse, low-entropy outputs, and how to push past the slop.
Oct 2025 · 10 min readComparison
GLM-4.5 vs Kimi-K2: Battle of the Agentic AI Giants
Z.ai's GLM-4.5 hybrid reasoning vs Moonshot AI's Kimi-K2 1T parameter architecture.
Aug 12, 2025 · 12 min readComparison
GLM-4.5 vs Qwen3-235B: The Ultimate Comparison
Z.ai's GLM-4.5 vs Alibaba's Qwen3-235B with massive parameter count and FP8 optimization.
Aug 12, 2025 · 15 min read