Union Alpha Is Pareto 26.9: Price, Benchmarks, Cost per Task
Union Alpha is Unbiased Pareto 26.9, a blended coding model at $2.50/$7.50 per Mtok that ties Astra on DeepSWE. Specs, benchmarks, cost.
Models & benchmarks
Leaderboards compress messy reality into a single number, then labs quote whichever number flatters the release. Suites saturate, contamination inflates scores, and the same model under different scaffolding can move ten points. What matters is which benchmarks still separate frontier models, and what each one refuses to measure.
- In this world
- 90 analyses
- Field tool
- Explore the value leaderboard ↗
Continue the descent.
Model releases, capability comparisons and rigorous analysis of the benchmarks used to rank them.
An Anthropic Researcher Quit. Safety Has a Price
Anthropic researcher Jacob Coxon quit on Sept 8, 2026, warning of runaway superintelligence. Days later the CEO called for a slowdown. What it costs builders.
Did ChatGPT Solve Navier-Stokes? What It Cost
No. OpenAI claims an internal model, not ChatGPT, found a Navier-Stokes proof. It is unverified. The 130B-token run lists near $6.5M.
Is GPT-6 Astra AGI? The 10 Tests That Decide It
OpenAI says the AGI era has started. Ten published tests, from ARC-AGI-3 to the market reaction, scored against the evidence with an honest tally.
Best Local Models for Hermes Agent in 2026
Hermes Agent rejects any model under 64,000 tokens of context. Which open-weight models clear that bar on 8GB, 16GB, 24GB and 32GB of VRAM.
GPT-6 Astra: Pricing, Benchmarks, Cost
GPT-6 Astra lists $10 and $50 per million tokens, 5x its own GPT-6 Sol tier. On Terminal-Bench it still solves a task for half of what Claude Fable 5.1 costs.
Gemini 3.8 Flash and Muse Spark 1.3 Pricing
Gemini 3.8 Flash costs $0.75 per million input tokens until December 31, then doubles. Muse Spark 1.3 is $1.25, or $0.10 if Meta can train on you.
Claude Fable 5.1: Pricing, Benchmarks, Cost
Claude Fable 5.1 holds Fable 5 rates at $10/$50 per million tokens and cuts the cache read to $0.25. Benchmarks, specs, and modeled cost per task.
New AI Models Released in September 2026: Prices
Every AI model released in September 2026, with dates, verified per-token prices and a primary source for each. Updated through the month as releases land.
Which AI Coding Model Burns the Fewest Tokens?
Independent measurement of eight frontier models: GPT-5 averages 2.38 file reads per task, Kimi-K2 averages 15.27, and neither can predict its own bill.
Muse Spark 1.2 Pricing: What a Task Really Costs
Muse Spark 1.2 at $1.25/$4.25 adds a $0.10/$0.20 contributor tier. Cost per task vs Claude Opus 5 and GPT-5.6 Sol, and what the data-for-discount really buys.
Terminal-Bench 4.0: The $234 Solved Task
Terminal-Bench 4.0 now runs 27 entries and GPT-6 Astra leads at 58.2 percent. Derived cost per solved task runs from 6 to 234 dollars, a 38x spread.
GLM-5.3-Flash: Price, Specs, and Benchmarks
GLM-5.3-Flash is Ox Alpha, officially: a 320B-A18B MoE model, 1M-token context, MIT license, and $0.15/$0.50 per million tokens. Specs, pricing, and benchmarks.
Ox Alpha Is Free: The Catch in the Data Policy
Ox Alpha lists $0 input and $0 output on OpenRouter and moves about 100 trillion tokens a day. Two data policies and an anonymous counterparty are the catch.
Qwen3.8-27B: What Free Open Weights Do to API Pricing
Apache 2.0 weights for Qwen3.8-27B landed on August 14, 2026. Strong vendor scores, hosted tokens at 80 percent off the flagship, and a catch in the local math.
Grok 4.6 Agent Cost: The 200K Price Cliff
Grok 4.6 kept the $2 and $6 headline but raised cached input 67 percent, and doubles every rate past 200K tokens. What that does to an agent bill.
AI Programming in 2026: Tools, Techniques, Evidence
AI writes 42 percent of committed code, yet the measured productivity gain stays contested. A sourced survey of the tools, techniques and evidence.
New AI Models Released in August 2026: The Full List
Every new AI model released in August 2026 with dates, verified per-token prices and primary sources, plus the price changes that landed alongside them.
Claude Opus 5.5 vs Sonnet 5: Which Model Should You Use?
Anthropic now starts most workloads on Claude Opus 5.5, at twice the Sonnet 5 token price. Compare cost per task, benchmarks, speed, and when Sonnet wins.
Prime Agent: The 95.5% Claim and the Exploit
Prime Intellect reports 95.5% on ARC-AGI-3. The ARC-verified ceiling on that same set is 30.2%. What the harness costs in tokens, and what the gap measures.
DeepSeek V4 Flash: Pricing, Benchmarks, and Cost
DeepSeek V4 Flash 0731 scores 50 on the Artificial Analysis Intelligence Index at $0.14 per million input tokens, ahead of the pricier V4 Pro at 44.
Why Local LLMs Fail at Agentic Coding in 2026
Identical open weights score 67.8 or 90.0 on SWE-bench depending on the harness. What the research says about where local coding agents break.
Claude Opus 5: Pricing, Benchmarks, and Cost
Claude Opus 5 ships at $5/$25 per million tokens with an effort dial, matching Fable 5 intelligence at half the price. Benchmarks, specs, and cost per task.
LLM Evaluation Metrics: Accuracy, Cost, and CI in 2026
A practitioner guide to LLM evaluation metrics: reference-based scores, LLM-as-judge, and human review, ranked by accuracy and cost per 10,000 responses.
What this topic covers
Model releases, capability comparisons and rigorous analysis of the benchmarks used to rank them.
This topic collects 90 analyses, and they are written to be read together rather than one at a time: the featured guide sets out the shape of the problem, and the pieces below work through the individual numbers, tradeoffs and edge cases behind it. Every figure is attributed to a primary source at the point it is used and carries the date it was verified, because prices and benchmark results in this area go stale in weeks rather than years. Where a number is modeled rather than measured, the assumptions are stated so the arithmetic can be checked.
The topics on this site overlap by design. This one runs into Coding agents and Local AI & hardware, and a question that starts in one usually ends in another: a pricing decision turns into a hardware decision, a benchmark result turns into a cost question. Follow the links inside the posts rather than treating these archives as separate shelves.
Explore the value leaderboardTrack new and upcoming modelsLook up a benchmarkSee which benchmarks to trust