The best open LLMs for your use case:

1Kimi K3Moonshot

Moonshot AI's 2.8-trillion-parameter open-weight frontier model with native vision, 1M-token context, and state-of-the-art agentic coding and tool use.

Speed:

Intelligence:

Price: (1M Tokens)

$3.00 / 15.00

Cached input: (1M Tokens)

$0.30

Context: (tokens)

1,048,576

Inputs:

ImageText

Benchmarks:

#1

MMMU-Pro

Multimodal - Vision

81.6
#1

GPQA-Diamond

General Knowledge

93.5
#1

HLE

General Knowledge

43.5
#1

Terminal-Bench 2.1

Coding Agents

88.3
#1

FrontierSWE

Coding Agents

81.2
#1

DeepSWE

Coding Agents

69
#1

Program Bench

Coding Agents

77.8
#1

MCP-Atlas

Agents and Function Calling

84.2
#1

MMMU-Pro

Multimodal - Vision

81.6
#1

GPQA-Diamond

General Knowledge

93.5
#1

HLE

General Knowledge

43.5
#1

Terminal-Bench 2.1

Coding Agents

88.3
#1

FrontierSWE

Coding Agents

81.2
#1

DeepSWE

Coding Agents

69
#1

Program Bench

Coding Agents

77.8
#1

MCP-Atlas

Agents and Function Calling

84.2
Try it out
2MiniMax-M3MiniMax

Next-generation reasoning model from MiniMax with frontier agentic, coding, and multimodal performance. Strong scores on SWE-Bench, BrowseComp, OmniDocBench, and IMO/USAMO competition reasoning.

Speed:

Intelligence:

Price: (1M Tokens)

$0.30 / 1.20

Cached input: (1M Tokens)

$0.06

Context: (tokens)

524,288

Inputs:

ImageText

Benchmarks:

#4

MMMU-Pro

Multimodal - Vision

78.1
#1

Video-MME v2

Multimodal - Vision

85.4
#1

Claw-Eval

Agents and Function Calling

74.5
#2

GPQA-Diamond

General Knowledge

92.9
#2

SWE-Bench Verified

Coding Agents

80.5
#2

SWE-Bench Pro

Coding Agents

59
#2

Apex Agents

Agents and Function Calling

27.7
#4

MMMU-Pro

Multimodal - Vision

78.1
#1

Video-MME v2

Multimodal - Vision

85.4
#1

Claw-Eval

Agents and Function Calling

74.5
#2

GPQA-Diamond

General Knowledge

92.9
#2

SWE-Bench Verified

Coding Agents

80.5
#2

SWE-Bench Pro

Coding Agents

59
#2

Apex Agents

Agents and Function Calling

27.7
Try it out

Use case:

Multimodal - Vision