The major LLMs in 2026

"Which LLM should I use" is the wrong first question. The right one is: closed or open, and at what size. Once you answer that, the shortlist writes itself.

This is post 2 of 7 in the Running series. The previous post sorted out where AI runs (cloud, local, edge). This one is the model tour: who the players are in 2026, what their numbers actually mean, and what each is good at.

The closed frontier

Closed models live behind APIs. You don't get the weights, you don't run them, you pay per token. In return you get the absolute frontier of capability.

Three labs matter:

• Anthropic, Claude. Opus 4.7 is the current heavyweight, around 57 on Artificial Analysis' Intelligence Index and leading SWE-bench Pro at 64.3%. Sonnet 4.6 is the daily driver: cheaper, fast, smart enough for almost everything. Haiku 4.5 is the small-fast tier.
• OpenAI, GPT. GPT-5.5 sits at the top of the leaderboard, 60 on Intelligence Index in xhigh-reasoning mode and 82.7% on Terminal-Bench 2.0. GPT-5.4 and GPT-5-mini fill out the tier below.
• Google, Gemini. Gemini 3.1 Pro is at 57 on Intelligence Index, leads scientific reasoning at 94.3% on GPQA Diamond, and ships with the longest production context window (1M tokens, experimental builds at 10M).

You will not run any of these on your laptop. Opus 4.7 needs eight to thirty-two H100s just to load the weights. The frontier is a datacenter phenomenon.

What "closed" actually buys you in 2026:

• Tool use that works on the first try. The frontier models are trained heavily on agentic loops, and it shows.
• Long-context reasoning that doesn't fall apart at 200K tokens.
• A pricing curve that drops every six months without you having to do anything.

What it costs you: privacy (every token leaves your machine), vendor lock-in (your prompts are tuned to one model's quirks), and a per-call bill that scales linearly with usage.

The open weights

Open models ship the weights. You can download them, run them, fine-tune them, embed them in a product. You don't need a permission slip from anyone.

The 2026 lineup, in rough order of how often you will actually hear them:

• Llama 4 (Meta). Maverick is 400B total / 17B active (Mixture-of-Experts), 10M-token context. Scout is the smaller variant. Llama is the default open model: best documented, best supported in every runtime.
• Qwen 3.5 (Alibaba). Ships dense (0.5B to 32B) and MoE (235B-A22B, 397B-A17B). Qwen punches well above its weight on reasoning and coding benchmarks. The dense small models are what you actually run on a laptop.
• DeepSeek V4 (DeepSeek). 1.6T total / 49B active, with Hybrid Compressed Sparse Attention that cuts inference FLOPs 27% vs V3.2 at 1M tokens. The reasoning leader on the open side. Each release moved the price floor for everyone.
• Mistral Large 3 (Mistral). 675B / 41B MoE, European, strong on European languages and code.
• Gemma 4 (Google). Smaller, dense, designed for on-device. The model you pick when you want open weights but don't want MoE complexity.
• Kimi K2.6 (Moonshot). Highest-ranked open-weights model on Intelligence Index right now (54). Worth knowing; not yet the default.

What "B" actually means

When you see "Llama 4 Maverick: 400B / 17B", the two numbers tell different stories.

The first is total parameters: every weight in the model file. That sets your storage and your VRAM floor. 400B parameters at FP16 is 800GB on disk; quantized to 4-bit it is around 200GB. Either way, you are not loading that on a 16GB laptop.

The second is active parameters: how many of those weights actually fire on any given token. Mixture-of-Experts models route each token through a small subset of the network, not all of it. 17B active means the cost-per-token math behaves like a 17B dense model, even though the file is 400B.

The rule of thumb:

• Total parameters set the VRAM floor (you have to fit the file).
• Active parameters set inference cost and speed (only those compute on each token).
• Dense models have total = active. MoE models split them.

This is why DeepSeek V4 at 1.6T total / 49B active runs on the same hardware as a much larger dense model would, but at a fraction of the per-token cost. MoE is the default at frontier scale in 2026 because it decouples size from cost.

Sizes that matter on a laptop

Frontier models live in datacenters. The ones you'll actually run locally are smaller. The useful tiers, with rough VRAM at 4-bit quantization:

• 1B to 3B. Tiny. Runs on anything, including a phone. Good for autocomplete, classification, structured extraction. Not for reasoning. (Gemma 4 1B, Qwen 3.5 0.5B/1.5B/3B.)
• 7B to 8B. The sweet spot for laptops. Around 5GB on disk, 6 to 8GB VRAM. Smart enough for most chat, drafting, summarization. (Llama 4 8B, Qwen 3.5 7B, Mistral 7B.)
• 13B to 32B. The serious local tier. 8 to 24GB VRAM. Noticeably better at multi-step reasoning. (Qwen 3.5 32B, Gemma 4 27B.)
• 70B class. Workstation territory. 40 to 50GB at 4-bit. Needs an M-series Mac with 64GB+ unified memory or a multi-GPU rig. (Llama 4 70B equivalent, DeepSeek-Distill 70B.)
• MoE flagships (200B to 1.6T total). Server territory. You don't run these unless you have a real machine.

What each is actually good at

Benchmarks are noisy. Here is the working version, from sitting with these models day to day:

• Coding agents and tool use. Claude Opus 4.7, then GPT-5.5, then DeepSeek V4 on the open side. Anthropic's lead on SWE-bench is real and you feel it in long agentic loops.
• Pure reasoning and math. GPT-5.5 in high-reasoning mode, Gemini 3.1 Pro for scientific topics, DeepSeek V4 if you want open.
• Cheap fast generation at volume. Haiku 4.5, GPT-5-mini, Qwen 3.5 7B locally. You are paying for throughput, not depth.
• Long context (1M+ tokens). Gemini 3.1 Pro is the production answer. Llama 4 Maverick is the open answer.
• Indian-language work. Qwen and Llama both do reasonable Hindi / Tamil / Bengali. Gemini handles Indic scripts cleanly. None of them are great at Hinglish (mixed-script) input — that's still a mess.

I would not pick a model based on a benchmark chart. I would pick two or three from the list above, run them on your actual workload for a week, and let the one that costs least to keep on staff win.

The pricing floor keeps falling

A useful frame: every six months, the model that was frontier-class becomes the price-floor model. Sonnet 4.6 today does what Opus 4 did eighteen months ago, at a fraction of the price. The same will be true of whatever Anthropic ships next.

The implication for anything you build: don't optimize your prompts for a specific model unless you have to. The model you target today will be obsolete by the time you ship; the model two tiers below it will probably be free.

What's next

This post sorted out who the models are. The next one zooms into the hardware you'd need to run them: what CPU, GPU, RAM, and VRAM each actually do for inference, and why VRAM is the bottleneck almost every time.