MODELS

Pick by use case, not by hype

Pricing, context, multimodal and positioning across mainstream overseas and Chinese models

19 models11 overseas8 Chinese

All models

Filter by region, expand for detail

19 models

Click the chevron to expand pricing, context and strengths. The model name links to the official site

Claude Fable 5Flagship

Anthropic$10 / $501M tokens

Anthropic's most capable widely released model for long-horizon agents

Input / 1M: $10
Output / 1M: $50
Context: 1M tokens
Modalities: ImgVidAud

Strengths

SOTA reasoning
Long-horizon agents
Code and math

Weaknesses

Top-tier pricing

Claude Opus 5Flagship

Anthropic$5 / $251M tokens

Latest Opus-tier model — complex agentic coding and enterprise work

Input / 1M: $5
Output / 1M: $25
Context: 1M tokens
Modalities: Img

Strengths

Complex agentic coding
Enterprise long-horizon tasks
1M context
Instruction following

Weaknesses

Top-tier pricing

Claude Sonnet 5Balanced

Anthropic$2 / $101M tokens

Latest Sonnet-tier model — high performance for coding and agents

Input / 1M: $2
Output / 1M: $10
Context: 1M tokens
Modalities: Img

Strengths

High-performance coding & agents
1M context
Cost-effective
Speed/intelligence balance

Weaknesses

Slightly behind Opus/Fable on hardest tasks

Claude Haiku 4.5Fast

Anthropic$1 / $5200K tokens

Fastest model with near-frontier intelligence for high-throughput

Input / 1M: $1
Output / 1M: $5
Context: 200K tokens
Modalities: Img

Strengths

Lowest latency
$1/$5 pricing
Near-frontier intelligence

Weaknesses

200K context
No adaptive thinking

GPT-5.6 SolFlagship

OpenAI$5 / $30400K tokens

OpenAI's current general flagship for complex reasoning, tool calls and multimodal

Input / 1M: $5
Output / 1M: $30
Context: 400K tokens
Modalities: ImgAud

Strengths

Complex reasoning
Stable tool calls
Multimodal
Largest ecosystem

Weaknesses

Needs proxy in China
Premium pricing

GPT-5.6 LunaFast

OpenAI$0.20 / $1.20128K tokens

OpenAI best value for high-volume everyday tasks (80% price cut, Aug 2026)

Input / 1M: $0.20
Output / 1M: $1.20
Context: 128K tokens
Modalities: Img

Strengths

Cheap
Fast
Full ecosystem

Weaknesses

Weaker at complex reasoning than GPT-5.6 Sol

Gemini 3.1 Pro PreviewFlagship

Google$2 / $121M tokens

Google's latest Pro preview, leading multimodal understanding, agents and vibe-coding

Input / 1M: $2
Output / 1M: $12
Context: 1M tokens
Modalities: ImgVidAud

Strengths

1M context
Native video/audio
Leading multimodal

Weaknesses

Preview release
Restricted access in China

Gemini 3.7 FlashFast

Google$0.75 / $3.751M tokens

Google's most capable Flash for agentic workflows and multimodal reasoning ($0.75/$3.75 promo through 2026, then $1.50/$7.50)

Input / 1M: $0.75
Output / 1M: $3.75
Context: 1M tokens
Modalities: ImgVidAud

Strengths

Agentic-workflow optimized
Multimodal reasoning
1M context
Free tier available

Weaknesses

Promo pricing reverts in 2027

Llama 4 MaverickOpen

MetaVariable1M tokens

Meta's open-source MoE flagship (400B total / 17B active), natively multimodal, self-hostable

Input / 1M: Variable
Output / 1M: Variable
Context: 1M tokens
Modalities: ImgVid

Strengths

Open-source, commercial
Natively multimodal
Self-hostable

Weaknesses

Requires inference infra

Grok 4.6Flagship

xAI$2 / $6500K tokens

xAI's flagship for code and everything else — agentic tool calling, minimal hallucinations, configurable reasoning

Input / 1M: $2
Output / 1M: $6
Context: 500K tokens
Modalities: Img

Strengths

Strong code & agent performance
Minimal hallucinations
Configurable reasoning effort
500K context

Weaknesses

Context behind the 1M tier
Younger ecosystem

GLM-5.3Flagship

智谱 AI待官方公布 / 待官方公布1M tokens

Zhipu's latest flagship (same base as 5.2, post-training boost) — coding +50%, emergent security capability; API pricing pending official launch

Input / 1M: 待官方公布
Output / 1M: 待官方公布
Context: 1M tokens
Modalities: Img

Strengths

Coding +50% vs 5.2 (Z.ai Code Bench)
CyberGym vulnerability-discovery SOTA
1M context / 128K max output
Terminal ops & long-horizon agent gains

Weaknesses

API not generally available yet
Text-only

DeepSeek V4 ProBalanced

DeepSeek$1.32 / $3.961M tokens

Open-source MoE flagship with cost-effective reasoning and thinking mode (official pricing now in USD)

Input / 1M: $1.32
Output / 1M: $3.96
Context: 1M tokens
Modalities: Img

Strengths

Open-source
1M context
Cost-effective reasoning
384K output

Weaknesses

Weak multimodal

DeepSeek V4 FlashFast

DeepSeek$0.44 / $1.321M tokens

DeepSeek's fast tier — $0.44/$1.32 peak, half off off-peak, built for volume

Input / 1M: $0.44
Output / 1M: $1.32
Context: 1M tokens
Modalities: Img

Strengths

Extremely low unit price
1M context
2500 concurrency limit
Thinking & non-thinking modes

Weaknesses

Shallower reasoning than V4 Pro

Qwen3.8 MaxFlagship

阿里云¥12 / ¥361M tokens

Alibaba flagship with thinking + non-thinking modes and image/video understanding

Input / 1M: ¥12
Output / 1M: ¥36
Context: 1M tokens
Modalities: ImgVid

Strengths

Thinking + non-thinking modes
Image/video understanding
1M context
128K output
Alibaba Cloud ecosystem

Weaknesses

Pricier than Chinese peers

Kimi K3Flagship

Moonshot AI¥20 / ¥1001M tokens

Kimi's flagship reasoning model — 1M context for long-horizon coding and end-to-end knowledge work

Input / 1M: ¥20
Output / 1M: ¥100
Context: 1M tokens
Modalities: Img

Strengths

1M context
Long-horizon coding
Always-on reasoning
Agent tool calls

Weaknesses

Top-tier pricing in China
Always-reasoning → higher latency

Doubao Seed 2.1 ProFlagship

字节跳动¥6 / ¥30256K tokens

Doubao's latest flagship (released 2026-06-23), rivaling overseas flagships across coding, agents and vision

Input / 1M: ¥6
Output / 1M: ¥30
Context: 256K tokens
Modalities: Img

Strengths

Flat pricing across context
Strong coding / agents
Low latency in China

Weaknesses

Pricier than Chinese peers

MiniMax-M3Flagship

MiniMax¥2.1 / ¥8.4512K tokens

MiniMax latest frontier model (released 2026-06), coding and agents

Input / 1M: ¥2.1
Output / 1M: ¥8.4
Context: 512K tokens
Modalities: Img

Strengths

Coding and agents
Long context
Frontier intelligence

Weaknesses

Newer ecosystem

MiMo-V2.5-ProOpen

小米Variable1M tokens

Xiaomi open-source flagship (1.02T total / 42B active MoE), MIT license, 1M context

Input / 1M: Variable
Output / 1M: Variable
Context: 1M tokens
Modalities: ImgAud

Strengths

MIT open-source, commercial
1T-param MoE
1M context
Strong agents

Weaknesses

Needs inference infra
No official API pricing

Fable 5.1FastNEW

AnthropicTBD待公布

New release — official data pending

Input / 1M: —
Output / 1M: —
Context: 待公布
Modalities: Img

Strengths

—

Weaknesses

—

Data sources and methodology

Anthropic — docs.anthropic.com models & pricing (official)
Google Gemini — ai.google.dev/gemini-api/docs/pricing (official)
OpenAI — developers.openai.com/api/docs/pricing (official)
xAI — docs.x.ai/developers/models (official, updated 2026-08-12)
DeepSeek — api-docs.deepseek.com pricing (official USD)
智谱 GLM — open.bigmodel.cn/pricing (official)
阿里云 Qwen — help.aliyun.com Model Studio pricing (official)
Moonshot Kimi — platform.kimi.com pricing (official)
字节 Doubao — volcengine.com Volcano Engine pricing (official)
MiniMax — platform.minimaxi.com pricing (official)
小米 MiMo — huggingface.co/XiaomiMiMo (open weights)
Prices are standard per-1M-token rates from public docs, with batch, cache and enterprise tiers cheaper

MODELS

Pick by use case, not by hype

Pricing, context, multimodal and positioning across mainstream overseas and Chinese models

19 models11 overseas8 Chinese

All models

Filter by region, expand for detail

19 models

Click the chevron to expand pricing, context and strengths. The model name links to the official site

Claude Fable 5Flagship

Anthropic$10 / $501M tokens

Anthropic's most capable widely released model for long-horizon agents

Input / 1M: $10
Output / 1M: $50
Context: 1M tokens
Modalities: ImgVidAud

Strengths

SOTA reasoning
Long-horizon agents
Code and math

Weaknesses

Top-tier pricing

Claude Opus 5Flagship

Anthropic$5 / $251M tokens

Latest Opus-tier model — complex agentic coding and enterprise work

Input / 1M: $5
Output / 1M: $25
Context: 1M tokens
Modalities: Img

Strengths

Complex agentic coding
Enterprise long-horizon tasks
1M context
Instruction following

Weaknesses

Top-tier pricing

Claude Sonnet 5Balanced

Anthropic$2 / $101M tokens

Latest Sonnet-tier model — high performance for coding and agents

Input / 1M: $2
Output / 1M: $10
Context: 1M tokens
Modalities: Img

Strengths

High-performance coding & agents
1M context
Cost-effective
Speed/intelligence balance

Weaknesses

Slightly behind Opus/Fable on hardest tasks

Claude Haiku 4.5Fast

Anthropic$1 / $5200K tokens

Fastest model with near-frontier intelligence for high-throughput

Input / 1M: $1
Output / 1M: $5
Context: 200K tokens
Modalities: Img

Strengths

Lowest latency
$1/$5 pricing
Near-frontier intelligence

Weaknesses

200K context
No adaptive thinking

GPT-5.6 SolFlagship

OpenAI$5 / $30400K tokens

OpenAI's current general flagship for complex reasoning, tool calls and multimodal

Input / 1M: $5
Output / 1M: $30
Context: 400K tokens
Modalities: ImgAud

Strengths

Complex reasoning
Stable tool calls
Multimodal
Largest ecosystem

Weaknesses

Needs proxy in China
Premium pricing

GPT-5.6 LunaFast

OpenAI$0.20 / $1.20128K tokens

OpenAI best value for high-volume everyday tasks (80% price cut, Aug 2026)

Input / 1M: $0.20
Output / 1M: $1.20
Context: 128K tokens
Modalities: Img

Strengths

Cheap
Fast
Full ecosystem

Weaknesses

Weaker at complex reasoning than GPT-5.6 Sol

Gemini 3.1 Pro PreviewFlagship

Google$2 / $121M tokens

Google's latest Pro preview, leading multimodal understanding, agents and vibe-coding

Input / 1M: $2
Output / 1M: $12
Context: 1M tokens
Modalities: ImgVidAud

Strengths

1M context
Native video/audio
Leading multimodal

Weaknesses

Preview release
Restricted access in China

Gemini 3.7 FlashFast

Google$0.75 / $3.751M tokens

Google's most capable Flash for agentic workflows and multimodal reasoning ($0.75/$3.75 promo through 2026, then $1.50/$7.50)

Input / 1M: $0.75
Output / 1M: $3.75
Context: 1M tokens
Modalities: ImgVidAud

Strengths

Agentic-workflow optimized
Multimodal reasoning
1M context
Free tier available

Weaknesses

Promo pricing reverts in 2027

Llama 4 MaverickOpen

MetaVariable1M tokens

Meta's open-source MoE flagship (400B total / 17B active), natively multimodal, self-hostable

Input / 1M: Variable
Output / 1M: Variable
Context: 1M tokens
Modalities: ImgVid

Strengths

Open-source, commercial
Natively multimodal
Self-hostable

Weaknesses

Requires inference infra

Grok 4.6Flagship

xAI$2 / $6500K tokens

xAI's flagship for code and everything else — agentic tool calling, minimal hallucinations, configurable reasoning

Input / 1M: $2
Output / 1M: $6
Context: 500K tokens
Modalities: Img

Strengths

Strong code & agent performance
Minimal hallucinations
Configurable reasoning effort
500K context

Weaknesses

Context behind the 1M tier
Younger ecosystem

GLM-5.3Flagship

智谱 AI待官方公布 / 待官方公布1M tokens

Zhipu's latest flagship (same base as 5.2, post-training boost) — coding +50%, emergent security capability; API pricing pending official launch

Input / 1M: 待官方公布
Output / 1M: 待官方公布
Context: 1M tokens
Modalities: Img

Strengths

Coding +50% vs 5.2 (Z.ai Code Bench)
CyberGym vulnerability-discovery SOTA
1M context / 128K max output
Terminal ops & long-horizon agent gains

Weaknesses

API not generally available yet
Text-only

DeepSeek V4 ProBalanced

DeepSeek$1.32 / $3.961M tokens

Open-source MoE flagship with cost-effective reasoning and thinking mode (official pricing now in USD)

Input / 1M: $1.32
Output / 1M: $3.96
Context: 1M tokens
Modalities: Img

Strengths

Open-source
1M context
Cost-effective reasoning
384K output

Weaknesses

Weak multimodal

DeepSeek V4 FlashFast

DeepSeek$0.44 / $1.321M tokens

DeepSeek's fast tier — $0.44/$1.32 peak, half off off-peak, built for volume

Input / 1M: $0.44
Output / 1M: $1.32
Context: 1M tokens
Modalities: Img

Strengths

Extremely low unit price
1M context
2500 concurrency limit
Thinking & non-thinking modes

Weaknesses

Shallower reasoning than V4 Pro

Qwen3.8 MaxFlagship

阿里云¥12 / ¥361M tokens

Alibaba flagship with thinking + non-thinking modes and image/video understanding

Input / 1M: ¥12
Output / 1M: ¥36
Context: 1M tokens
Modalities: ImgVid

Strengths

Thinking + non-thinking modes
Image/video understanding
1M context
128K output
Alibaba Cloud ecosystem

Weaknesses

Pricier than Chinese peers

Kimi K3Flagship

Moonshot AI¥20 / ¥1001M tokens

Kimi's flagship reasoning model — 1M context for long-horizon coding and end-to-end knowledge work

Input / 1M: ¥20
Output / 1M: ¥100
Context: 1M tokens
Modalities: Img

Strengths

1M context
Long-horizon coding
Always-on reasoning
Agent tool calls

Weaknesses

Top-tier pricing in China
Always-reasoning → higher latency

Doubao Seed 2.1 ProFlagship

字节跳动¥6 / ¥30256K tokens

Doubao's latest flagship (released 2026-06-23), rivaling overseas flagships across coding, agents and vision

Input / 1M: ¥6
Output / 1M: ¥30
Context: 256K tokens
Modalities: Img

Strengths

Flat pricing across context
Strong coding / agents
Low latency in China

Weaknesses

Pricier than Chinese peers

MiniMax-M3Flagship

MiniMax¥2.1 / ¥8.4512K tokens

MiniMax latest frontier model (released 2026-06), coding and agents

Input / 1M: ¥2.1
Output / 1M: ¥8.4
Context: 512K tokens
Modalities: Img

Strengths

Coding and agents
Long context
Frontier intelligence

Weaknesses

Newer ecosystem

MiMo-V2.5-ProOpen

小米Variable1M tokens

Xiaomi open-source flagship (1.02T total / 42B active MoE), MIT license, 1M context

Input / 1M: Variable
Output / 1M: Variable
Context: 1M tokens
Modalities: ImgAud

Strengths

MIT open-source, commercial
1T-param MoE
1M context
Strong agents

Weaknesses

Needs inference infra
No official API pricing

Fable 5.1FastNEW

AnthropicTBD待公布

New release — official data pending

Input / 1M: —
Output / 1M: —
Context: 待公布
Modalities: Img

Strengths

—

Weaknesses

—

Data sources and methodology

Anthropic — docs.anthropic.com models & pricing (official)
Google Gemini — ai.google.dev/gemini-api/docs/pricing (official)
OpenAI — developers.openai.com/api/docs/pricing (official)
xAI — docs.x.ai/developers/models (official, updated 2026-08-12)
DeepSeek — api-docs.deepseek.com pricing (official USD)
智谱 GLM — open.bigmodel.cn/pricing (official)
阿里云 Qwen — help.aliyun.com Model Studio pricing (official)
Moonshot Kimi — platform.kimi.com pricing (official)
字节 Doubao — volcengine.com Volcano Engine pricing (official)
MiniMax — platform.minimaxi.com pricing (official)
小米 MiMo — huggingface.co/XiaomiMiMo (open weights)
Prices are standard per-1M-token rates from public docs, with batch, cache and enterprise tiers cheaper