The best large language models (LLMs) in 2026
The best large language models (LLMs) right now are Claude Opus 5.5, Claude Sonnet 5.5 and GPT-6.1 Sol, out of 23 we have reviewed. Each is ranked by our rating, with its price, our verdict and who it suits.
Updated Oct 23, 2026 · Ranked from our reviews · How we review
- 1
Anthropic's new default earns the title: first on Artificial Analysis's independent index, and a fifth cheaper per token than the Opus it replaces.
Best for: Teams running long coding agents, code audits or document-heavy research where finishing the task matters more than the cost of each call.
Price: API: $4 per 1M input tokens, $20 per 1M output; cache reads $0.20, 5-minute cache writes $5; Batch API 50% off; fast mode $8/$40; US-only inference 1.1x.
Read our Claude Opus 5.5 review4.7/5 - 2
Within two points of Opus 5.5 on Anthropic's knowledge-work test at half the price, which turns Opus into a deliberate upgrade rather than the default.
Best for: Teams running coding agents, support automation or document work at volume who want near-Opus quality at half the token price.
Price: API: $2 per 1M input tokens, $10 per 1M output tokens; cache reads $0.10; Batch API 50% off.
Read our Claude Sonnet 5.5 review4.6/5 - 3
For most OpenAI customers this is the flagship that matters: a point behind Astra on Artificial Analysis's index at a fifth of the token price.
Best for: Teams building agents and business workflows on OpenAI who want near-Astra results at mid-tier cost and can live with slower output.
Price: API: $2 per 1M input tokens, $10 per 1M output; cached input $0.10; Batch/Flex half price; prompts over 272K input tokens pay 2x input and 1.5x output for the whole request.
Read our GPT-6.1 Sol review4.6/5 - 4
Haiku 5.5 outscores Gemini 3.8 Flash on independent tests at a sixth of the cost per task, which resets what anyone should expect from a budget model.
Best for: High-volume classification, extraction, support and subagent work on short prompts.
Price: API: $0.10 input / $0.50 output per 1M tokens for prompts up to 100K tokens; $0.50 / $2.50 above that; cache reads $0.01; Batch API 50% off.
Read our Claude Haiku 5.5 review4.5/5 - 5
A 27B model with an Apache licence that, on Alibaba's own numbers, edges past Claude Opus 4.6 Max at coding and fits on a single 24 GB card at 4-bit.
Best for: Developers who want the strongest permissively licensed model that fits on one high-end GPU, especially for coding and computer-use agents.
Price: Free weights under Apache 2.0. Qwen Cloud: $0.50 input / $3.00 output per 1M tokens ($0.10 cached). Third-party hosts on OpenRouter from about $0.04 / $1.25.
Read our Qwen3.8-27B review4.5/5 - 6
GLM-5.3-Flash gets most of GLM-5.3's coding ability for a tenth of the token price, and its MIT licence leaves nothing for a lawyer to read.
Best for: Teams running high-volume coding or multimodal agents who want an open model with no licence conditions and very low API costs.
Price: Open weights under MIT. Z.ai API: $0.15 input / $0.50 output per 1M tokens ($0.03 cached); FlashX $0.37 / $1.25.
Read our GLM-5.3-Flash review4.3/5 - 7
Luna scores five points below Claude Haiku 5.5 on Artificial Analysis's index and costs a third as much per task, which is the whole decision in one line.
Best for: OpenAI-based apps doing classification, extraction or chat at very high volume.
Price: API: $0.10 per 1M input tokens, $0.50 per 1M output tokens; cached input $0.01; long prompts over 272K cost more.
Read our GPT-6 Luna review4.3/5 - 8
Near-frontier coding-agent scores for $0.15 per million input tokens, provided you can accept China-based data handling or host a 763B-parameter checkpoint yourself.
Best for: Cost-sensitive teams running coding and automation agents who can use a China-hosted API or a third-party host, or have multi-GPU servers for the open weights.
Price: Free MIT-licensed weights. DeepSeek API: $0.15 input / $0.60 output per 1M tokens off-peak, $0.30 / $1.20 at weekday peak; cache hits $0.003 to $0.006.
Read our DeepSeek-V4.1-Flash review4.2/5 - 9
Strong agent and coding scores for 6 billion active parameters, under a licence that is stricter than the one on Qwen's own flagship.
Best for: Teams running internal coding or agent workloads that outgrow a 27B model, and developers who want a very cheap hosted multimodal API.
Price: Open weights under the Qwen Community License 1.0 (separate licence for hosting and AI-assistant businesses). Qwen3.8-Flash API: $0.15 input / $0.47 output per 1M tokens; cache hits $0.016.
Read our Qwen3.8-Flash-Next review4.2/5 - 10
Quick to stream and broad in what it reads, but Haiku 5.5 and GPT-6 Luna now charge a third of its input price and a fifth of its output.
Best for: Teams on Google Cloud running high-volume extraction or audio and video understanding where time to first token does not matter.
Price: API: $0.30 input / $2.50 output per 1M tokens; cached input $0.03; Batch and Flex $0.15 / $1.25; free tier (prompts may be used for training).
Read our Gemini 3.5 Flash-Lite review4.1/5 - 11
Broad inputs and a free tier keep Gemini 3.8 Flash useful, but it now spends more per task than rivals that score higher, and its price doubles in January.
Best for: Teams processing audio, video and long PDFs inside Google's ecosystem who can tolerate slow first responses.
Price: API: $0.75 input / $3.75 output per 1M tokens until 31 Dec 2026, then $1.50 / $7.50; Batch and Flex half price; free tier with limits.
Read our Gemini 3.8 Flash review4.1/5 - 12
Glimmer is the open model most carefully fitted to the graphics card people own, and Qwen3.8-27B beat it on the benchmarks within weeks.
Best for: Developers building local, always-on agents on one consumer GPU who want a permissive licence from a US lab.
Price: Free weights under Apache 2.0; no Meta API. Hosted on OpenRouter from $0.30 per 1M input and $1.20 per 1M output tokens (DeepInfra).
Read our Muse Glimmer review4.1/5 - 13
Wonder costs what Claude Opus 5.5 costs and wins only five of the fourteen rows Meta chose to publish against it.
Best for: Teams already on Meta's API that need a stronger model for long, multi-file coding or legal document work than Muse Spark provides.
Price: API: $4 input, $0.20 cached input, $20 output per 1M tokens; no Contributor mode. Wonder Full only in Muse Code Power ($50/month) and Maximum ($200/month).
Read our Muse Wonder review4.1/5 - 14
Cheap, long-context and well-mannered in agent loops, but squeezed from above by Muse Wonder and from the side by an open Spark at the same price.
Best for: Teams that want a cheap, long-context model for sub-agents and tool loops and are content to send prompts to Meta's API.
Price: Standard: $1.25 input, $0.15 cached input, $4.25 output per 1M tokens. Contributor mode: $0.10 / $0.002 / $0.20, and Meta may train on your content.
Read our Muse Spark 1.3 review4.0/5 - 15
A serious open-weight agent model with unusually strong safety results, held back by hosted prices that match closed Spark and a licence that asks for your birth date.
Best for: Organisations that need a capable agent model on their own hardware, with stronger refusal behaviour than other open models, and can live with Meta's attribution clause.
Price: Free weights under the Muse Model License (attribution required; $20M revenue threshold for model-as-a-service). Hosted on OpenRouter at $1.25 input, $0.15 cached, $4.25 output per 1M tokens.
Read our Muse Spark Open review4.0/5 - 16
Strong for its price on long, mixed-media inputs, held back by a licence that makes lawyers read twice.
Best for: Cost-sensitive teams processing long documents or video who want an open model and can accept its licence terms.
Price: Open weights under the MiniMax community licence. API per 1M tokens on prompts up to 512K: $0.30 input, $1.20 output, $0.06 cached; double above 512K; priority tier 1.5x.
Read our MiniMax-M3 review3.9/5 - 17
Europe finally has a trillion-parameter model of its own; independent tests place it in the upper middle of the pack, not at the frontier.
Best for: European organisations that need a frontier-class model with data sovereignty and a path to self-hosting.
Price: Preview API: $1.36 per 1M input tokens ($0.14 cached), $4.18 per 1M output; batch 50% off. Open weights promised by end of October 2026.
Read our Mistral Large 4 review3.9/5 - 18
Apache 2.0 weights, two-H100 hosting and every EU language make Command A+ easy to approve, but Cohere still publishes no benchmark scores to judge it by.
Best for: Enterprises building private, multilingual RAG and agent systems who want a permissively licensed model they can host on two GPUs.
Price: Free weights under Apache 2.0. Cohere API rate not published; OpenRouter lists $0.30 input / $1.50 output per 1M tokens. Private deployment by quote.
Read our Command A+ review3.8/5 - 19
Cheap on the rate card and dear in practice: Grok 4.7 thinks for so long that its cost per task lands among the highest measured.
Best for: Developers who need live X search inside an agent and can tolerate long thinking times, or teams already committed to SpaceXSI's platform.
Price: API: $2 input / $6 output per 1M tokens under 200K-token prompts, $4 / $12 at or above; cached input $0.50; US endpoint 10% extra.
Read our Grok 4.7 review3.8/5 - 20
A capable four-GPU coding model whose licence excludes large companies and whose API price now loses to Mistral's own Large 4.
Best for: Companies under $20 million in monthly revenue that want to self-host a strong coding and agent model on one four-GPU server.
Price: Open weights under a modified MIT licence (no rights for companies above $20M monthly revenue). API: $1.50 per 1M input tokens, $7.50 per 1M output; batch 50% off.
Read our Mistral Medium 3.5 review3.8/5 - 21
Cheap, fast and cleanly licensed, but independent tests put it well behind the open models it shares a server with.
Best for: Teams wanting a low-cost, Apache-licensed model for chat, documents and images, on Mistral's API or their own GPUs.
Price: Free weights under Apache 2.0. API: $0.15 input / $0.60 output per 1M tokens; batch 50% off; cached input up to 90% off.
Read our Mistral Small 4 review3.7/5 - 22
Still a capable Apache 2.0 workhorse, but Alibaba now sells the same model hosted for less than any third party charges to run the open weights.
Best for: Teams that need the largest Apache 2.0 Qwen on their own hardware and already have a multi-GPU server or a 256 GB machine to run it.
Price: Free weights under Apache 2.0; Qwen3.5-Plus on Alibaba Cloud and third-party hosting billed separately.
Read our Qwen3.5-397B-A17B review3.6/5 - 23
Xing4.0 is proof that China can train a useful agent model without NVIDIA, and a reasonable cheap coder, but not a reason to leave Qwen.
Best for: Teams on Huawei Ascend hardware, or those wanting a small Apache-licensed agent model to fine-tune for domain tasks.
Price: Free weights under Apache 2.0. No first-party API; OpenRouter lists a free endpoint (hosted by Phala) under its free-model request caps.
Read our Xing4.0-29B-A4B review3.6/5