Your unbiased guide to the world's smartest AIs
One-click access to today's frontier leaders. Ranked by capability, updated weekly.
First Mythos-class model cleared for general use. Leads LMArena text overall at 1508 Elo with 95% on SWE-bench Verified and top GDPval-AA knowledge-work scores — gains widen on long-horizon agentic runs.
Codebase-scale agentic coding: parallel-subagent dynamic workflows migrate hundreds of thousands of lines from kickoff to merge. 88.6% on SWE-bench Verified, same price as 4.7.
Dominates reasoning and multimodal tasks. Scored 77% on ARC-AGI-2 — double its predecessor — and leads on graduate-level science benchmarks.
xAI's recommended chat and code model: 83.3% on Terminal-Bench 2.1, 4.2× fewer output tokens than Opus 4.8 on SWE Bench Pro, served at 80 TPS. $2/$6 with 500K context.
Meta's first paid frontier API: native multi-agent orchestration, MCP and computer use, 1M self-managed context. #6 on LMArena text overall at 1487 Elo — cheapest agentic option helloai tracks at $1.25/$4.25.
Moonshot's 2.8T MoE flagship (16 of 896 experts active): natively multimodal, 1M context, long-horizon coding and knowledge work. API at $3/$15 with $0.30 cache-hit input; LMArena text overall ~1486 Elo.
Elo ratings from Chatbot Arena blind votes. These shift weekly — here's the current snapshot.
No API key, no usage limits, no data leaving your machine. The best open-weight models ranked by Elo.
Google's Gemma 4 31B ranks #3 among all 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.
Alibaba's Qwen3 flagship dense model. Matches Qwen2.5-72B performance in a 32B package that fits a single RTX 4090, with hybrid thinking/non-thinking modes and Apache 2.0.
A 30B mixture-of-experts with only ~3B parameters active per token. We measured 44.7 tok/s on a two-GPU budget cluster — faster than the dense 8B while packing 4x the capacity. Apache 2.0.
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.
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 19.3 tok/s — 12x faster than single-card CPU offload. Apache 2.0.
The lowest-friction way into real local AI: fits any 6 GB+ GPU at Q4_K_M and measured 43.5 tok/s on our budget two-GPU cluster. Hybrid thinking modes, Apache 2.0.
No hype. Where each model actually leads, based on benchmarks and real-world usage as of today.
Highest Elo in the tracked set at 1508 on LMArena text overall. Mythos-class capability with conservative safety fallbacks to Opus 4.8 on flagged queries.
95.0% on SWE-bench Verified and 80% on SWE-bench Pro — well ahead of Opus 4.8. Best pick for long-horizon agentic coding when the 2× price premium is justified.
Leads on PhD-level benchmarks like GPQA and ARC-AGI subsets. Claude and Grok are strong contenders.
xAI's default chat model — direct answers without corporate polish, now with coding-grade speed at 80 TPS and 4.2× token efficiency on engineering tasks.
Weekly analysis, honest takes, and hidden gems. No engagement bait.
Moonshot's 2.8T Kimi K3 lands on the public API at $3/$15 with 1M context. helloai drops GPT-5.6 Sol and adds Moonshot under the six-model cap.
OpenAI and Anthropic are no longer fighting only on LMArena — ChatGPT Work and Claude Cowork compete for the surface where finished work ships.
Meta ships Muse Spark 1.1 with a public API at $1.25/$4.25. helloai drops Qwen3.7-Max and adds Meta as its budget agentic pick at 1487 Elo.
xAI launched Grok 4.5 on July 8 — 83.3% on Terminal-Bench 2.1, 4.2× token efficiency, and $2/$6 pricing. helloai's tracked Grok entry moves from 4.3 to 4.5.
OpenAI ends the partner-only preview on July 9. GPT-5.6 Sol replaces GPT-5.5 in helloai's tracked set at the same $5/$30 rate — the upgrade path teams waited on since June 26.
helloai's leaderboard shows nominal per-token rates. Prompt caching and batch APIs stack underneath — turning a $5/MTok flagship into $0.25 on repeated context.
One command routes a full agent harness to MIT-licensed weights at Ollama Pro prices. That stack does not kill frontier labs, but it is eating their easiest revenue.
ARC-AGI-3 launched March 25 with humans at 100% and frontier AI at 0.51%. GPT-5.5 and Opus 4.7 barely move the needle — exposing a gap arena Elo cannot see.
Fable 5 returns globally July 1 after an 18-day suspension; GPT-5.6 Sol ships only to vetted partners. June 2026 made frontier availability a policy variable, not just an engineering milestone.
xAI's frontier model is now GA on Amazon Bedrock with the same $1.25/$2.50 pricing, 1M context, and configurable reasoning — the cheapest tracked frontier model just became enterprise-deployable.