The Best LLMs in 2026: A Plain-English Comparison

Two years ago, picking an AI model meant choosing between a handful of names. Today there are dozens, a new one lands most weeks, and every launch post claims the crown.

This guide is for the version of you that doesn’t want to spend an afternoon reading leaderboards. It answers three questions per model - what is it good at, what does it cost, what’s the catch - and nothing else. No computer science degree required.

We’re MindsHub, by MindsDB. Our agents call every model on this page all day to get real work done, so keeping score is our job rather than a hobby. And we don’t sell a model, which is why we can rank them honestly.

Every price and ranking on this page was re-checked against Artificial Analysis’s full table on August 27, 2026. In this field that date is part of the fact - and this refresh moved a lot: OpenAI cut GPT-5.6 Sol to $4 / $20, Grok 4.6 slipped a point and got dearer per task, and five models we had been estimating now have published measurements.

The short answer

• One model for everything → Claude Sonnet 5 or GPT-5.6 Terra. Both draft, summarize, analyze and code well enough that you’ll rarely hit a wall.
• The hard stuff → Claude Opus 5, which has led the public intelligence rankings since July. GLM-5.3 is three points behind for under a third of the cost per task if your budget is the constraint.
• Big PDFs, images, audio or video → Gemini 3.7 Flash. It handles mixed-media inputs and a million-token context at one of the lowest prices in the table.
• High volume, low budget → GPT-5.6 Luna, the cheapest model in the table by measured task cost. If you want more capability for very little more, GLM-5.3-Flash scores five points higher at $0.09 per task with MIT weights, as long as nothing is waiting on the answer. DeepSeek V4 Flash is the low-cost option if you need open weights and speed together.
• Coding on a budget → GLM-5.3, a few points off the frontier at $1.40 / $4.40. If it has to run on your own hardware, Kimi K3 is the strongest model you can download today.

Then do the one thing no leaderboard can do for you: run the same real task through two of them and keep the one whose answer you’d actually send.

If that’s all you needed, you’re done. The rest explains the why, and why most teams have quietly stopped picking just one.

What each model is for, and what it really costs

Grouped by maker, in rough order of overall adoption. This is not a ranking - don’t read the top row as a winner. Let the Best for column and the two cost columns decide.

Two columns need a word of explanation, because they’re the ones that save you the reading:

• Quality is a plain-English tier, with Artificial Analysis’s composite score underneath for anyone who wants the number. Frontier is the top of the board, near-frontier is within a few points, strong handles most professional work, solid is fine for routine jobs.
• Per task is what it actually costs to finish one piece of work, not per million tokens. It’s the number the price column fails to predict, and the two disagree wildly. We dug into why in a companion piece.

The fine print

• A token is about ¾ of a word, so the million tokens most of these models read at once is roughly 750,000 words, or a long book. Grok 4.6 stops at 500,000. Claude’s current generation is the exception on density: it uses a tokenizer that bills about 30% more tokens for the same text, so a million tokens there is closer to 575,000 words.
• Consumer apps are priced differently. ChatGPT, Claude and Gemini subscriptions are a flat monthly fee. The prices above are what developers and agents pay per token.
• ^† Tiered on long prompts. Grok 4.6 goes to $4 / $12 above ~200,000 tokens, and the whole GPT-5.6 line roughly doubles above 272,000. In each case crossing the line re-prices the entire request, so budget long-document work at the upper tier. Claude, Kimi, Qwen, GLM, DeepSeek and the Gemini Flash line are flat all the way up.
• ^‡ Promotional. Gemini 3.7 Flash is $0.75 / $3.75 through December 31, 2026. Google has not published the rate that follows.
• Open weights have no single price. The DeepSeek rows follow the current first-party provider and rate shown by Artificial Analysis. Another host, or your own hardware, can produce a different bill for the same weights.
• ^* Licences differ more than the tags suggest. GLM-5.3 and Qwen3.8-Max are proprietary today, though the previous GLM-5.2 and the newer GLM-5.3-Flash are both open - Flash under MIT. Kimi K3 is downloadable under Moonshot’s own license, DeepSeek V4 uses MIT, and the separate Qwen3.8 2.4T A95B checkpoint is open weights under a commercially restricted Qwen license. Read the terms before reselling access.

New this month, and whether it changes your pick

Several recent releases should move your pick.

• Grok 4.6 (August 12) is three points off the top and one off GPT-5.6 Sol, at $2 / $6 - a third of Sol’s output price, but a fifth more per finished task, and its own top effort setting scores worse than one notch down. Only for the live web.
• Gemini 3.7 Flash (August 13) is four points better than 3.6 Flash and much stronger at code. Google put 3.7 on a $0.75 / $3.75 promotional rate until the end of the year. If you run volume through Gemini, this one pays for itself. Yes.
• GLM-5.3 (August 18) matches Kimi K3’s score at $1.40 / $4.40 and costs less per task than either Grok 4.6 or GPT-5.6 Sol. Artificial Analysis still classifies it as proprietary, so for now it’s an API. Yes, if renting is fine.
• Qwen split its flagship in two. The proprietary, multimodal Qwen3.8-Max API scores 58. The downloadable Qwen3.8 2.4T A95B checkpoint matches that score, but it is text-only and its license restricts commercial use. Yes, if that tradeoff fits.
• Muse Glimmer and Nemotron 3.5 Lightning (August 10 and 11) are both 30-billion-parameter open agent models built to run locally. Glimmer scores 35 at $0.06 a task under Apache 2.0; Nemotron scores 24 - the lowest figure we track - but streams at 300 tokens a second, which is the trade it was built to make. Only for on-device work, or for agent steps you have already broken down.
• GLM-5.3-Flash is the release we’d look at first if you are paying for volume: 57 on the index, MIT weights, images in, and $0.09 per task. It is slow, and for batch work that costs you nothing. Yes, unless a person is waiting.
• Inkling (July 15) is the first production model from Mira Murati’s Thinking Machines Lab, released under Apache 2.0 - text, image and audio in, 975 billion parameters, and the leading open-weights release from a US lab on Artificial Analysis’s ranking. It scores 42, which several cheaper open models beat, so the licence is the reason to take it. Inkling Small landed one point behind it on well under a third of the parameters, twice as fast and far cheaper. Yes, if the licence is what you need.
• Motif 3 is Korea’s sovereign open model - 314 billion parameters, free to download, scoring 47. The licence is research-only, so it is an evaluation and fine-tuning story rather than a production one. Not yet, unless you’re researching.

How to read this without a spreadsheet

• The test isn’t your job. A model that aces graduate physics can still write clunky emails. Try two or three on a real task you do every week; that beats any leaderboard.
• The top of the board is crowded. Six models sit inside three points - Claude Opus 5 at 63, Claude Fable 5 at 62, GPT-5.6 Sol at 61, and Grok 4.6, Kimi K3 and GLM-5.3 tied at 60 on a 0–100 scale - which is close enough to be noise. When quality is that level, cost and speed decide.
• Headline numbers are the expensive setting. Leaderboards list each model once per reasoning effort - how long it may think before answering - and labs submit at maximum. Claude Opus 5’s cheapest setting uses about an eighth of the tokens its priciest one does. That dial will save you more money than switching models.

Want the deeper checklist for production use? We wrote 12 things to weigh when choosing an LLM. Otherwise, the rundown below is enough.

The models, one by one

GPT-5.6 (OpenAI) - competent at everything, at three price points

GPT is the name most people recognize, because it powers ChatGPT. July’s generation dropped the old mini/nano naming for three tiers: Sol is the flagship - cut to $4 / $20 in late August - and runs a dead heat with Claude Opus 5 at the top of the public coding-agent rankings. Terra at $2 / $12 is the sensible default. Luna at $0.20 / $1.20 is the bargain of the table, and it now scores alongside good open models while costing less to rent than most cost to host.

OpenAI’s lineup is also the most honest about its own price ladder: across a seventeen-fold price range from Luna to Sol, what you pay per task tracks what you pay per token closely - and since the August cuts, Luna and Terra both come in slightly cheaper per task than their rate cards imply. That is rarer than it sounds.

Pick it when you want one provider that’s competent at everything. Skip it when your work is dominated by huge documents, or when price is the binding constraint.

Claude (Anthropic) - hard reasoning, and work someone signs off on

Claude Opus 5 has led the public intelligence rankings since late July at $5 / $25, which is half of Claude Fable 5 - a model Anthropic’s own documentation still calls its most capable. They sit one point apart, so hold that distinction lightly and let the price decide. Claude Sonnet 5 is the everyday tier at $2 / $10, and that rate is now permanent: on August 10 Anthropic canceled the increase to $3 / $15 it had scheduled for September.

Two things are unusual here. Claude reads its full million-token context at the standard rate, where Grok and GPT-5.6 re-price long prompts upward. And every Claude model in the current generation costs more per finished task than its sticker price implies, because these models are built to keep chewing on a problem - the previous-generation Haiku 4.5 is the exception, and it barely reasons. On genuinely hard work that’s what you want. On routine work you’re paying for thoroughness you didn’t need.

Claude has a reputation for writing that doesn’t sound robotic and for being careful about getting things right, which makes it a favorite in legal, financial and other detail-heavy work. Claude also powers many of the most popular AI coding tools, which is why it ranks where it does here despite a smaller consumer audience than ChatGPT.

Pick it when the task is hard or the output carries risk. Skip it when you’re paying by the token for routine work.

If you see Claude Mythos 5 in a benchmark table, it’s the same underlying model as Fable 5 with fewer safeguards, available by invitation to security researchers and government cyber defenders. You can’t buy it. Claude Opus 4.8 is still active and supported, with no retirement date announced.

Gemini (Google) - fast multimodal work at volume

Gemini is natively multimodal, which is the technical way of saying it reads text, images, audio and video in the same request. Gemini 3.7 Flash is the current pick for that work: it reads a million tokens, handles long documents and media, and costs $0.75 / $3.75 through the end of 2026.

Shipped August 13, three weeks after 3.6 Flash, 3.7 is four points better on the composite score and much stronger on code. Google put 3.7 on a $0.75 / $3.75 promotional rate through the end of 2026. Gemini 3.5 Flash-Lite covers the truly high-throughput end at $0.30 / $2.50.

The heavier Gemini 3.5 Pro keeps slipping - Google says it’s still testing with partners, and reporting in July put it months behind schedule. Don’t plan around it.

Pick it when your inputs are large or mixed-media, or you live in Google Workspace. Skip it when you need frontier-level reasoning or weights you can host yourself.

Grok (SpaceXAI) - cheap per token, expensive per task

Grok is wired into X and the live web, so it can answer questions about this morning’s news. That is the reason to pick it, and as of this refresh it is close to the only one. Grok 4.6 scores 60 at $2 / $6 against Sol’s $4 / $20 - a third of the output price - and yet costs more per finished task than Sol does, $1.23 against $1.01. It thinks its way past the discount. It ties the two GLM-5.3 releases for the best GDPval result of any non-Anthropic model - the test built around realistic professional deliverables rather than puzzles - behind only Claude Opus 5.

There is a stranger detail, and it is the most useful thing on this page if you run Grok. Its reasoning dial stops paying off before the top: at high it scores 61 for $0.94 a task, and at xhigh it scores 60 for $1.23. Turning it all the way up costs a third more and scores a point worse. Check where your own dial sits.

Two other caveats. Test it yourself on terminal work if that’s your use case: published results put it either level with the frontier or well behind it, depending on which version of the shell-driven benchmark you read. And its context window is 500,000 tokens, half of what most of this table offers, with the rate doubling above 200,000. Its predecessor Grok 4.5 is four points back at $0.43 a task, which now makes it the better-value half of the pair.

Pick it when you need live information in the same call as the reasoning. Skip it when the bill is what you are optimising - the price page will mislead you here more than anywhere else in this table.

GLM (Zhipu) - frontier-class coding without frontier bills

GLM, from Beijing-based Zhipu, is the dark-horse story of 2026. GLM-5.3 landed on August 18 and scores level with Kimi K3 at $1.40 / $4.40. Its gains come from post-training rather than a larger base model. Its headline result is a security one: on a benchmark that scores whether a model can spot real vulnerabilities in source code, Zhipu reports it narrowly ahead of both Anthropic’s and OpenAI’s frontier models.

One asterisk matters, and it’s the reason GLM-5.3 isn’t tagged “open” in the table above. Zhipu has not released the weights, so for now this is an API you rent rather than a model you can host. It would tie Kimi K3 at the top of the open-weights board if that changes. The previous release, GLM-5.2, now is listed as open - 53 at $1.40 / $4.40 - so the older model is the one you can host.

Then Zhipu did something more interesting. GLM-5.3-Flash is the first natively multimodal model in the GLM-5 line, it takes images as well as text, its weights are out under MIT, and it scores 57 at $0.15 / $0.50 - a ninth of what GLM-5.2 costs, for four points more. At $0.09 per task it is the cheapest thing on our board anywhere near that score, and the catch is patience rather than money: Artificial Analysis measures it as slow and very verbose, so a finished answer takes noticeably longer than the price suggests. 320 billion parameters in total, 18 billion active per token.

Pick it when you want strong coding at a low price. Skip it when you need multimodal input or weights you can host today - or, for Flash specifically, when someone is waiting on the answer.

Kimi (Moonshot AI) - the strongest model you can run yourself

Kimi is built for agentic software engineering: jobs where the model plans, writes, tests and fixes across many steps. Kimi K3 ties GLM-5.3’s score and leads the models whose weights are actually available. It reads a million tokens, handles text and images, and posts some of the best terminal and coding-agent numbers anyone publishes. Its weights have been out since late July, which makes it the strongest model you can genuinely run yourself today.

Two honest notes. At $3 / $15 it’s priced like a mid-tier closed model, not a budget one - the era of dirt-cheap Chinese frontier models is ending. And the license is Moonshot’s own rather than MIT or Apache: self-hosting and fine-tuning are fine, reselling access at scale needs a read.

Pick it when you want the best downloadable model and can live with a custom license. Skip it when the API bill is the point - GLM is cheaper for similar work.

Qwen (Alibaba) - a multimodal API and a separate open checkpoint

Qwen is one of the most prolific families in AI. Qwen3.8-Max is the proprietary flagship API: it supports text, images and video, reads a million tokens, scores 58 and costs $2 / $6. Artificial Analysis measures it at $0.91 per task and 21 output tokens per second - the slowest streaming rate in this table, so this is a capable multimodal option rather than a speed or cost leader.

The downloadable counterpart is Qwen3.8 2.4T A95B, released August 12. It also scores 58, but it is text-only and comes under a commercially restricted Qwen license. Artificial Analysis measures it at $0.81 per task with a 984,000-token context.

Pick Qwen3.8-Max when you want a capable multimodal API with a million-token context. Pick the 2.4T checkpoint when you need the weights and can accept text-only input and the license. Skip both when speed or a permissive license is the priority.

DeepSeek V4 - low-cost models with weights you can run

DeepSeek proved you don’t need a frontier-sized budget to get near-frontier results. DeepSeek V4 is open-weight under MIT, in two flavors: the full-strength V4 Pro and the cheaper, faster V4 Flash. Artificial Analysis currently lists Flash at $0.44 / $1.32 and $0.11 per task, while Pro is $1.32 / $3.96 and $0.27 per task. Flash is the sensible default for high-volume work; Pro is there when a task needs headroom.

Where you rent open weights still matters. Artificial Analysis reports one provider entry per row, while another host or your own hardware can produce a different bill for the same model.

Pick it when volume is high and budget is tight. Skip it when you need the last few points of quality, or multimodal input.

Muse (Meta) - cheap and capable, if Meta is acceptable

Meta spent years as the champion of open-weight AI with Llama, then changed course: the Muse family is closed, with no downloadable weights. Muse Spark 1.2 is a capable agent-focused model at $1.25 / $4.25 that reads a million tokens and handles every input type including audio. It also powers the free Meta AI across WhatsApp, Instagram and Facebook, which arguably makes it the most widely deployed model on this list.

Watch one line on its price sheet: a “contributor” tier at $0.10 / $0.20, roughly a fifteenth of standard, in exchange for letting Meta train on your prompts and the model’s replies. Fair trade for hobby projects, easy no for anything confidential.

Meta has also started shipping small open models again - Muse Glimmer, a 30-billion-parameter agent model under Apache 2.0 that runs on a single consumer GPU. It scores 35 at $0.06 a task and takes images as well as text, with a 131,000-token window - the shortest here. Not a frontier model, but a genuinely useful local one.

Pick it when you want frontier-adjacent quality at open-model prices. Skip it when you need weights, or your data can’t go to Meta.

Also worth knowing

• MiniMax-M3 - a quietly strong multimodal all-rounder at $0.30 / $1.20 with a million-token context. Its weights are available, but the MiniMax Community License restricts commercial use.
• Nemotron (NVIDIA) - the most completely open releases on this page: weights, training data, recipe and RL environment. The 30-billion-parameter Nemotron 3.5 Lightning rents for about $0.07 / $0.22, which is close to free, and streams at 300 tokens a second. Fast, though not the fastest here: Gemini 3.7 Flash runs at 362 and Gemini 3.5 Flash-Lite at 386. It scores 24, so treat it as a worker for steps you have already specified rather than a model that will work anything out for you.
• Inkling and Inkling Small (Thinking Machines Lab) - Apache 2.0 releases from the lab founded by OpenAI’s former CTO, and the cleanest licence on this page: no commercial catch, no field-of-use clause, weights on Hugging Face. Both take text, images and audio. Neither leads on score - 42 and 41 - and the lab has been refreshingly direct that they aren’t the strongest models available. Take them for the terms, not the ranking. Small is the one to reach for: one point behind on well under a third of the parameters, twice as fast, and $0.07 a task against $0.34.
• Motif 3 (Moreh) - Korea’s sovereign open model, 314 billion parameters with 13 billion active, scoring 47 and free to download. Two catches: the licence is non-commercial research only, and it is very verbose, generating well over twice the median output on the same benchmark suite. Nobody publishes a cost per task for it, because there is no priced endpoint to measure.
• Llama (Meta) - the family that made local AI mainstream, now in maintenance mode. Still downloadable and widely supported, no longer where the action is.
• Mistral (France) - Europe’s flagship lab, focused on small, efficient models that are cheap to run and friendly to data-residency rules. A sensible pick for EU teams and on-device work.

Open vs closed: what actually matters

You’ll see models split into “closed” (you rent access through an API) and “open” (the weights are published, so you can download and run them). The camps flipped in 2026: Meta went closed with Muse, while strong downloadable models now come from Moonshot, DeepSeek, MiniMax and Alibaba’s separate open checkpoint. Three questions decide it for you:

• Where does your data go? Closed means your prompts travel to the provider. Fine for most work; not fine for patient records, unreleased financials or legal matters.
• What does volume cost? Closed frontier models add up fast at scale. Open models are dramatically cheaper for routine work, and free beyond hardware if you host them.
• Are you locked in? Build everything on one provider and you inherit its price changes and its deprecations. With open weights, you can change host, or stop renting altogether.

The honest answer for 2026 is that you want both: closed frontier models for the hard 10%, open models for the high-volume 90%.

You don’t have to pick one

The best model for one task is rarely the best model for the next. Drafting an email and untangling a quarter of messy financials are different jobs, and paying frontier prices for both is like taking a sports car to do the grocery run.

That matters more once AI works as an agent - software that plans, uses tools and grinds through a task over many steps. Agents burn far more text than a chat does, because they read, act, check and retry. Run the routine steps on a cheap open model, save the frontier model for the hard part, and you get the same result for a fraction of the cost.

Agents are also where the price column stops predicting your invoice. Some of the cheapest models here take so many turns to finish a job that most of their discount disappears, and one of the pricier ones works out cheaper than the budget option it should have beaten. Cost per task is the number that shows it.

Running the right engine for each job is what we built MindsHub around. MindsHub Cowork is one workspace where you hand over a whole task - “pull last quarter’s refunds, explain the biggest movers, build me a dashboard” - and collect the finished work rather than a chat transcript. Underneath, Unified Inference lets you switch among the models available there from a dropdown, with no per-provider API keys to juggle. Your agent, history and memory carry over when you switch.

That’s what being model-neutral means. The model is a setting, not a life sentence, which is also why we can compare these models honestly - we don’t have a horse in the race.

MindsHub Cowork is free, with no subscription on either tier. Bring keys you already have and pay your provider directly, or skip keys and run our MindsHub Air model on five million tokens a month. Pro is pay as you go at the published per-model rates. Or browse the use-case gallery to see what people hand off.

Frequently asked questions

What’s the best LLM in 2026? There’s no single winner, and the top is closer than ever. Claude Opus 5 leads the public intelligence rankings at 63, with Claude Fable 5 at 62, GPT-5.6 Sol at 61 and Grok 4.6, Kimi K3 and GLM-5.3 all at 60 - six models inside three points, close enough to be noise. “Best” depends on the job: Gemini 3.7 Flash is the value pick for multimodal work, Grok 4.6 is the one to pick for live information, and GLM-5.3, DeepSeek and GLM-5.3-Flash win on cost.

What’s the best free or open-source model? Kimi K3, today. It scores 60, level with GLM-5.3, and unlike the proprietary GLM model you can download it. Qwen3.8 2.4T A95B scores 58, but its license carries commercial restrictions. For permissive weights, DeepSeek V4 uses MIT and remains the budget favorite.

Which LLM is the cheapest? GPT-5.6 Luna has the lowest measured task cost in this table at $0.05, with a $0.20 / $1.20 rate card. DeepSeek V4 Flash is the cheapest open-weight option in the table above at $0.44 / $1.32 and $0.11 per measured task; further down the page, Inkling Small ($0.07 a task, Apache 2.0), Nemotron 3.5 Lightning ($0.08) and GLM-5.3-Flash ($0.09, MIT) go cheaper still. Gemini 3.7 Flash is the cheapest capable audio-and-video model here at $0.75 / $3.75.

Is GPT better than Claude? It’s a photo finish that changes with each release. Claude leads today: Opus 5 sits at the top of the intelligence rankings with GPT-5.6 Sol two points back, while the two are within half a point of each other at the top of the coding-agent rankings - and after August’s cuts Sol is the cheaper of the two on input, $4 against $5. At the everyday tier - GPT-5.6 Terra against Claude Sonnet 5 - they’re close enough that price and style should decide. Note that Claude models cost more per finished task than their sticker price suggests, and OpenAI’s don’t.

What’s the best LLM for coding? Claude Opus 5 and GPT-5.6 Sol sit within half a point of each other at the top of the public coding-agent rankings. Open models have closed the gap fast: GLM-5.3 and Kimi K3 post frontier-class coding results at a fraction of the price, which is why they’re popular for high-volume development.

Do I have to commit to one model? No, and you probably shouldn’t. Unified Inference lets you route each kind of work to the model that suits it - cheap open models for routine steps, frontier models for the hard parts - and you keep your work when you switch.

How often does this change? Constantly. Several models on this page launched this month, and prices have already moved. That’s the real argument for staying flexible rather than betting everything on one model. We re-check this page every couple of weeks; between visits, the Artificial Analysis leaderboard is the best place to watch the board move.

---

MindsHub by MindsDB puts every major model behind one endpoint. Unified Inference serves frontier and open models through the OpenAI and Anthropic request formats on one API key and one bill, and MindsHub Cowork is the agent workspace on top of that catalog - delegate entire projects and collect finished, shareable results, with work running on interchangeable open-source agent harnesses, Anton and Hermes. Founded 2018 in Berkeley. Backed by Benchmark, Mayfield, Y Combinator, and NVIDIA.