CanMyPCRunAILocal AI compatibility

Frontier models · snapshot 13 October 2026

Which AI model is best at what?

The most capable AI models right now, ranked on one independent benchmark and then split by task, because no single model wins everything. Every number links to where it came from.

Best at each job

Overall intelligence

Claude Opus 5.5

Highest Artificial Analysis Intelligence Index score at 58. Claude Sonnet 5.5 is next at 56; GPT-6 Astra, Claude Fable 5.1 and Gemini 4 Argon are level at 53.

Independent measurement

Agentic coding

Claude Sonnet 5.5

Artificial Analysis measured 64% on Terminal-Bench 4.0, ahead of roughly 60% for Opus 5.5 and GPT-6 Astra. Anthropic reports 70.6% with its launch setup, so the exact result depends on the harness.

Independent measurement

Office and knowledge work

Claude Opus 5.5

GDPval-AA v2.1, which scores real professional deliverables: 1846 Elo against 1735 for Fable 5.1 and 1542 for GPT-6 Astra.

Vendor-reported benchmark

Business automation

GPT-6 Astra

AutomationBench: 41.4%, narrowly ahead of Opus 5.5 at 40.0%. One of the few results in Anthropic’s own table that Anthropic does not win.

Vendor-reported benchmark

Lowest measured cost near the top

GPT-6.1 Sol

Scores 52, one point behind GPT-6 Astra and Gemini 4 Argon. Artificial Analysis measured $0.72 per index task; that is benchmark-specific, but lower than the other selected releases scoring 50 or more.

Independent measurement

Speed and price

Gemini 3.8 Flash

141.4 output tokens per second at $0.75 / $3.75 per million tokens: the cheapest of these releases and marginally faster than Claude Sonnet 5.5 (135.2), but with a lower index score (41). Measured speeds drift from week to week.

Independent measurement

Best model you can download

MiMo-V2.6-Pro

Highest-scoring open-weights model at 46, ahead of GLM-5.3 (45) and Kimi K3 (44). At about one trillion parameters it needs a server, not a desktop PC.

Independent measurement

Best downloadable model for a home PC

Qwen3.8 27B

Scores 34, the highest of the open models under 100 billion parameters on Artificial Analysis’s list. At 27 billion parameters it is small enough that Perplexity’s local agent runs it on a 24 GB graphics card.

Independent measurement

Selected leading releases

Artificial Analysis Intelligence Index v4.3.2 combines ten evaluations — coding, reasoning, science, knowledge work and tool use — which Artificial Analysis runs itself. Each model is shown at its highest-effort setting. Prices are US dollars per million input / output tokens. This is a maintained selection of current frontier and leading open-weights releases, not a copy of the complete leaderboard. Mistral Large 4 is hosted only for now: Mistral has promised its weights by the end of October but had not named a licence when this was checked.

#ModelReleasedIndexPriceSpeedWeights
1Claude Opus 5.5Anthropic22 Sep 202658$4 / $2095.2 tokens/sHosted only
2Claude Sonnet 5.5Anthropic28 Sep 202656$2 / $10135.2 tokens/sHosted only
3Claude Fable 5.1Anthropic1 Sep 202653$10 / $5068.3 tokens/sHosted only
3GPT-6 AstraOpenAI3 Sep 202653$10 / $5062.5 tokens/sHosted only
3Gemini 4 ArgonGoogle30 Sep 202653$2 / $10 (introductory)Not yet measuredHosted only
6GPT-6.1 SolOpenAI29 Sep 202652$2 / $1064.8 tokens/sHosted only
7MiMo-V2.6-ProXiaomi—46——Downloadable
7Grok 4.7SpaceXAI21 Sep 202646$2 / $683.8 tokens/sHosted only
9GLM-5.3Z AI—45——Downloadable
10Kimi K3Moonshot AI (Kimi)—44——Downloadable
11Gemini 3.8 FlashGoogle2 Sep 202641$0.75 / $3.75141.4 tokens/sHosted only
12Mistral Large 4Mistral AI6 Oct 202638$1.36 / $4.18Not reportedHosted only
13Qwen3.8 27BAlibaba—34——Downloadable

“—” means the figure is not published on the source page, not that it is zero. Scores from different index versions are not comparable, so this table uses only v4.3.2. Gemini 4 Argon’s price is introductory: Google says it rises to $4 / $20, and has not said when. At launch on 30 September its access was limited to cyber defenders in Google’s Fairwind programme, with no public API or Gemini app release date announced, even though Artificial Analysis has scored it.

Head to head: Anthropic’s launch figures

From Anthropic’s Opus 5.5 announcement. These are vendor-reported: the company releasing the model chose the tests and ran them. Treat them as a claim to check against independent results, not as a verdict.

BenchmarkOpus 5.5Fable 5.1GPT-6 AstraOpus 5
Terminal-Bench 4.0Agentic coding in a terminal66.4%55.8%57.9%52.3%
FrontierCode v1.1Hard software tasks54.4%50.3%53.3%48.0%
GDPval-AA v2.1Professional deliverables (Elo)1846173515421708
AutomationBenchBusiness workflows40.0%31.4%41.4%26.9%
Humanity’s Last ExamExpert reasoning, with tools67.7%65.6%57.2%63.6%
OSWorld 2.0Operating a computer81.8%80.7%Not reported74.0%

What this means if you run AI on your own PC

None of the top three can be downloaded; they only run on their makers’ servers. The best open-weights models are close behind, but at 750 billion to 2.8 trillion parameters they need data-centre hardware. What runs on a home graphics card is smaller models from the same families — and whether a given one fits depends on your video memory, not on this ranking.

Scan your PCBrowse models you can runGrok Bot, Muse and Dots explainedLatest AI news

Sources

All figures retrieved 13 October 2026. Leaderboards change as models are added and re-tested.