The best open LLMs for your use case:

1Kimi K3.1 ProMoonshot

Moonshot AI's flagship open-weight 2.8T-parameter multimodal MoE, post-trained from Kimi K3 for long-horizon coding, agentic work, and stronger factual accuracy across a 1M-token context.

Speed:

Intelligence:

Price: (1M Tokens)

$2.00 / 10.00

Cached input: (1M Tokens)

$0.20

Context: (tokens)

1,048,576

Inputs:

ImageText

Benchmarks:

#1

GPQA-Diamond

General Knowledge

94.2
#1

HLE

General Knowledge

51.1
#1

MMMU-Pro

Multimodal - Vision

84.1
#1

DeepSWE

Coding Agents

75.5
#1

Kimi Code Bench v2

Coding Agents

77.1
#1

GDPval-AA

Agents and Function Calling

1846
#1

Apex Agents

Agents and Function Calling

69.4
#2

Program Bench

Coding Agents

31.5
#1

GPQA-Diamond

General Knowledge

94.2
#1

HLE

General Knowledge

51.1
#1

MMMU-Pro

Multimodal - Vision

84.1
#1

DeepSWE

Coding Agents

75.5
#1

Kimi Code Bench v2

Coding Agents

77.1
#1

GDPval-AA

Agents and Function Calling

1846
#1

Apex Agents

Agents and Function Calling

69.4
#2

Program Bench

Coding Agents

31.5
Try it out
2MiniMax M3.1 FlashMiniMax

MiniMax's natively multimodal 428B-A23B MoE built for fast, low-cost coding, agentic workflows, and computer use across a 1M-token context.

Speed:

Intelligence:

Price: (1M Tokens)

$0.15 / 0.60

Cached input: (1M Tokens)

$0.015

Context: (tokens)

1,048,576

Inputs:

ImageText

Benchmarks:

#3

HLE

General Knowledge

43.4
#3

DeepSWE

Coding Agents

73.7
#3

HLE

General Knowledge

43.4
#3

DeepSWE

Coding Agents

73.7
Try it out

Use case:

Personal Agents