Updated November 17, 2026

Hello, Ai

Your unbiased guide to the world's smartest AIs

Find the right AI. Introduce it to another.
See what you can make together.

Or catch up on the latest AI dispatches →
00 Capability mapCurated strengths · select a model
Curated strengths of each tracked model, by category
ModelOverall PreferenceCoding & EngineeringHard Reasoning & ScienceHonest Daily Use
Curated strengthNot listedCurated strengthNot listed
Not listedCurated strengthNot listedNot listed
Not listedCurated strengthNot listedCurated strength
Not listedCurated strengthNot listedNot listed
Not listedCurated strengthNot listedNot listed
Not listedCurated strengthNot listedNot listed

Strengths are curated and not exclusive. Category leaders are named in the spec plate and under Category Breakdown.

Spec plateGemini 4 Argon · Google
Arena Elo
1522 ± 6
USD per 1M tokens
$2/M in · $10/M out
Context
1M ctx
Snapshot
Nov 11, 2026
LMArena text-overall
Category leader in
Overall Preference · Hard Reasoning & Science

Different minds. Shared possibilities. Start a conversation in our app ↗

Models tracked
6
Elo snapshot
Nov 11, 2026
Data updated
November 17, 2026
Rated / not yet rated
4 / 2

This week's frontier

6 frontier models we track, chosen by us. Filter by task or price. Updated weekly.

1
Intro $2/$10
Google

Gemini 4 Argon

Google's public Gemini 4 model, id gemini-4-argon. Introductory paid-tier price is $2/$10 through November 30, 2026, then $4/$20, with no free tier and a 1,048,576 token context. Rendered-board write-ups still mark Gemini 4 Carbon Preliminary, so Carbon is not this card.

$2/M in$10/M out1M ctx
2
40% Cheaper
Anthropic

Claude Opus 5.5

Anthropic's September 22 successor to Opus 5 at $4/$20 with 1M context and cache reads at $0.20. Anthropic says it matches Fable 5.1 on most work and costs about 40% less to run than Opus 5. Text Arena lists claude-opus-5.5-high, so this card's Elo is the model's own score. On the November 15 Code Arena WebDev board, claude-opus-5.5-max is second, behind claude-fable-5.2-xhigh.

$4/M in$20/M out1M ctx
3
Budget Agentic
Meta

Muse Spark 1.3

Meta's Muse Spark 1.3 — agentic and coding upgrade over 1.2, using ~20% fewer tool calls and ~25% fewer tokens in Meta's own comparisons. Max reasoning is live (arena muse-spark-1.3-max). Cheapest tracked API at $1.25/$4.25 with 1M context.

$1.25/M in$4.25/M out1M ctx
4
Agentic Code
xAI

Grok 4.7

xAI's recommended model for chat and code. $2/$6 under 200K prompt tokens and $4/$12 at or above 200K, with a 500K window. Text Arena lists grok-4.7-xhigh at 1443 on 22,219 votes. Code Arena WebDev lists it at 1640 on 8,488 votes.

$2/M in$6/M out500K ctx

Not yet rated

These models have no Elo of their own on the board yet. They are listed here, unranked, and never count toward a best match.

This week's ranking

LMArena text Elo from the arena.ai board of Nov 11, 2026, with list price and context. Models without a score of their own are listed below the ranking, unranked.

Run it yourself

No API key, no usage limits, no data leaving your machine. The best open-weight models ranked by Elo.

1
Efficiency Champion
Google DeepMind

Gemma 4 31B

Google's Gemma 4 31B ranks among the top open-weight models on LMArena with an Elo of 1452 — beating models 20x its size. Fits a single RTX 4090 at Q4_K_M and ships Apache 2.0.

18 GB VRAM50 t/sApache 2.0
RTX 4090 (24 GB)
2
Coding Champion
Qwen Team

Qwen3.8-27B

Alibaba's Qwen3.8 generation dense flagship, replacing the Qwen3 32B card at a smaller footprint: 27B parameters, 262K native context, and hybrid tool-use. Fits a single RTX 4090 at Q4 (~15.3 GiB weights); community-reported 65 tok/s decode. Apache 2.0.

17 GB VRAM65 t/sApache 2.0
RTX 4090 (24 GB)
3
Single-GPU Pick
Mistral AI

Mistral Small 3.2 24B

The strongest model that comfortably fits a single 24 GB GPU. Adds vision and tool-use over 3.1, tightly instruction-tuned, and Apache 2.0 with no usage restrictions — a dependable all-rounder for local daily use.

14 GB VRAM24.2 t/sApache 2.0⚡ Independently measured
RTX 5060 8GB + RTX 5060 8GB (llama.cpp RPC split, 2.5GbE)
4
MoE Champion
Qwen Team

Qwen3 30B-A3B

A 30B mixture-of-experts with only ~3B parameters active per token. We measured 53.8 tok/s on a two-GPU budget cluster — faster than the dense 14B while packing twice the capacity. Apache 2.0.

14 GB VRAM53.8 t/sApache 2.0⚡ Independently measured
RTX 5060 8GB + RTX 5060 8GB (llama.cpp RPC split, 2.5GbE)
5
OpenAI Local
OpenAI

gpt-oss-20b

OpenAI's Apache 2.0 MoE (~21B total / ~3.6B active). Native MXFP4 fits ~16 GB; we measured 109.17 tok/s on the budget two-GPU cluster — the fastest open-weight card we track. Arena ~1318 Elo.

16 GB VRAM109.17 t/sApache 2.0⚡ Independently measured
RTX 5060 8GB + RTX 5060 8GB (llama.cpp RPC split, 2.5GbE)
6
Two-GPU Pick
Qwen Team

Qwen3 14B

The model that makes pooling two budget GPUs worth it: ~10.5 GB of weights fit neither of our 8 GB cards alone, but the cluster ran it at a measured 37.1 tok/s — far faster than spilling to CPU. Apache 2.0.

11 GB VRAM37.1 t/sApache 2.0⚡ Independently measured
RTX 5060 8GB + RTX 5060 8GB (llama.cpp RPC split, 2.5GbE)

Where each one leads

Category leaders from the tracked set, as of this week's snapshot.

Dispatches from the frontier

Weekly analysis, honest takes, and hidden gems. No engagement bait.