Comparisons of Medium Open Source AI Models (40B-150B)
Open source AI models with between 40B to 150B parameters. Models are considered Open Source (also commonly referred to as open weights) where their weights are accessible to download. This allows self-hosting on your own infrastructure and enables customizing the model such as through fine-tuning. Click on any model to see detailed metrics. For more details including relating to our methodology, see our FAQs.
gpt-oss-120B (high) and 
Qwen3-Coder-Next are the highest intelligence Medium open source models, defined as those with 40B-150B parameters, followed by Qwen3 Next 80B A3B & 
K2 Think V2.
Highlights
Openness
Artificial Analysis Openness Index: Results
Intelligence
Artificial Analysis Intelligence Index
Artificial Analysis Intelligence Index v4.0 includes: GDPval-AA, 𝜏²-Bench Telecom, Terminal-Bench Hard, SciCode, AA-LCR, AA-Omniscience, IFBench, Humanity's Last Exam, GPQA Diamond, CritPt. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.
{"@context":"https://schema.org","@type":"Dataset","name":"Artificial Analysis Intelligence Index","creator":{"@type":"Organization","name":"Artificial Analysis","url":"https://artificialanalysis.ai"},"description":"Artificial Analysis Intelligence Index v4.0 incorporates 10 evaluations: GDPval-AA, 𝜏²-Bench Telecom, Terminal-Bench Hard, SciCode, AA-LCR, AA-Omniscience, IFBench, Humanity's Last Exam, GPQA Diamond, CritPt","measurementTechnique":"Independent test run by Artificial Analysis on dedicated hardware.","spatialCoverage":"Worldwide","keywords":["analytics","llm","AI","benchmark","model","gpt","claude"],"license":"https://creativecommons.org/licenses/by/4.0/","isAccessibleForFree":true,"citation":"Artificial Analysis (2025). LLM benchmarks dataset. https://artificialanalysis.ai","data":""}
Intelligence Evaluations
While model intelligence generally translates across use cases, specific evaluations may be more relevant for certain use cases.
Artificial Analysis Intelligence Index v4.0 includes: GDPval-AA, 𝜏²-Bench Telecom, Terminal-Bench Hard, SciCode, AA-LCR, AA-Omniscience, IFBench, Humanity's Last Exam, GPQA Diamond, CritPt. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.
Model Size: Total and Active Parameters
The total number of trainable weights and biases in the model, expressed in billions. These parameters are learned during training and determine the model's ability to process and generate responses.
The number of parameters actually executed during each inference forward pass, expressed in billions. For Mixture of Experts (MoE) models, a routing mechanism selects a subset of experts per token, resulting in fewer active than total parameters. Dense models use all parameters, so active equals total.
Intelligence vs. Active Parameters
Artificial Analysis Intelligence Index v4.0 includes: GDPval-AA, 𝜏²-Bench Telecom, Terminal-Bench Hard, SciCode, AA-LCR, AA-Omniscience, IFBench, Humanity's Last Exam, GPQA Diamond, CritPt. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.
The number of parameters actually executed during each inference forward pass, expressed in billions. For Mixture of Experts (MoE) models, a routing mechanism selects a subset of experts per token, resulting in fewer active than total parameters. Dense models use all parameters, so active equals total.
Intelligence vs. Total Parameters
Artificial Analysis Intelligence Index v4.0 includes: GDPval-AA, 𝜏²-Bench Telecom, Terminal-Bench Hard, SciCode, AA-LCR, AA-Omniscience, IFBench, Humanity's Last Exam, GPQA Diamond, CritPt. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.
The total number of trainable weights and biases in the model, expressed in billions. These parameters are learned during training and determine the model's ability to process and generate responses.
Context Window
Context Window
Larger context windows are relevant to RAG (Retrieval Augmented Generation) LLM workflows which typically involve reasoning and information retrieval of large amounts of data.
Maximum number of combined input & output tokens. Output tokens commonly have a significantly lower limit (varied by model).
{"@context":"https://schema.org","@type":"Dataset","name":"Context Window","creator":{"@type":"Organization","name":"Artificial Analysis","url":"https://artificialanalysis.ai"},"description":"Context Window: Tokens Limit; Higher is better","measurementTechnique":"Independent test run by Artificial Analysis on dedicated hardware.","spatialCoverage":"Worldwide","keywords":["analytics","llm","AI","benchmark","model","gpt","claude"],"license":"https://creativecommons.org/licenses/by/4.0/","isAccessibleForFree":true,"citation":"Artificial Analysis (2025). LLM benchmarks dataset. https://artificialanalysis.ai","data":""}
| Weights | Provider Benchmarks | |||||||
|---|---|---|---|---|---|---|---|---|
gpt-oss-120B (high) OpenAI | 33 | 117B (5.1B active at inference time) | 131k | $0.3 | 338 | 🤗 |    +20 more | View |
Qwen3-Coder-Next Alibaba | 28 | 79.7B (3B active at inference time) | 256k | $0.5 | 105 | 🤗 |   | View |
Qwen3 Next 80B A3B (Reasoning) Alibaba | 26 | 80B (3B active at inference time) | 262k | $1.9 | 167 | 🤗 |  +4 more | View |
K2 Think V2 MBZUAI Institute of Foundation Models | 25 | 70B | 262k | - | - | Not available | - | View |
gpt-oss-120B (low) OpenAI | 24 | 117B (5.1B active at inference time) | 131k | $0.3 | 340 | 🤗 |   +17 more | View |
 GLM-4.6V (Reasoning) Z AI | 23 | 108B | 128k | $0.5 | 85 | 🤗 |   +1 more | View |
GLM-4.5-Air Z AI | 23 | 106B (12B active at inference time) | 128k | $0.4 | 105 | 🤗 | +1 more | View |
 INTELLECT-3 Prime Intellect | 22 | 107B | 131k | $0.4 | 84 | 🤗 | | View |
 Devstral 2 Mistral | 22 | 125B | 256k | - | 74 | 🤗 | | View |
K2-V2 (high) MBZUAI Institute of Foundation Models | 21 | 70B | 512k | - | - | 🤗 | - | View |
 Ring-flash-2.0 InclusionAI | 21 | 103B (6.1B active at inference time) | 128k | $0.2 | 97 | 🤗 | | View |
 Hermes 4 - Llama-3.1 70B (Reasoning) Nous Research | 20 | 70.6B | 128k | $0.2 | 85 | 🤗 | | View |
Qwen3 Next 80B A3B Instruct Alibaba | 20 | 80B (3B active at inference time) | 262k | $0.9 | 175 | 🤗 |  +4 more | View |
K2-V2 (medium) MBZUAI Institute of Foundation Models | 19 | 70B | 512k | - | - | 🤗 | - | View |
 Llama Nemotron Super 49B v1.5 (Reasoning) NVIDIA | 19 | 49B | 128k | $0.2 | 76 | 🤗 | | View |
Llama 3.3 Nemotron Super 49B v1 (Reasoning) NVIDIA | 18 | 49B | 128k | - | - | 🤗 | - | View |
GLM-4.6V (Non-reasoning) Z AI | 17 | 108B | 128k | $0.5 | 60 | 🤗 | | View |
 DeepSeek R1 Distill Llama 70B DeepSeek | 16 | 70B | 128k | $0.9 | 41 | 🤗 |  | View |
Ling-flash-2.0 InclusionAI | 15 | 103B (6.1B active at inference time) | 128k | $0.2 | 73 | 🤗 | | View |
Llama Nemotron Super 49B v1.5 (Non-reasoning) NVIDIA | 15 | 49B | 128k | $0.2 | 74 | 🤗 | | View |
K2-V2 (low) MBZUAI Institute of Foundation Models | 14 | 70B | 512k | - | - | 🤗 | - | View |
 Kimi Linear 48B A3B Instruct Kimi | 14 | 49.1B (3B active at inference time) | 1.00M | - | - | 🤗 | - | View |
Llama 3.3 Nemotron Super 49B v1 (Non-reasoning) NVIDIA | 14 | 49B | 128k | - | - | 🤗 | - | View |
 Llama 3.3 Instruct 70B Meta | 14 | 70B | 128k | $0.7 | 134 | 🤗 |  +19 more | View |
Hermes 4 - Llama-3.1 70B (Non-reasoning) Nous Research | 14 | 70.6B | 128k | $0.2 | 74 | 🤗 | | View |
Llama 4 Scout Meta | 13 | 109B (17B active at inference time) | 10.0M | $0.3 | 124 | 🤗 |  +7 more | View |
 Command A Cohere | 13 | 111B | 256k | $4.4 | 66 | 🤗 |  | View |
Llama 3.1 Nemotron Instruct 70B NVIDIA | 13 | 70B | 128k | $1.2 | 42 | 🤗 | | View |
Llama 3.2 Instruct 90B (Vision) Meta | 12 | 90B | 128k | $0.7 | 37 | 🤗 |  +1 more | View |
 Jamba 1.7 Mini AI21 Labs | 7 | 52B (12B active at inference time) | 258k | $0.3 | - | 🤗 | | View |