Compare at a glance.

Anthropic · version 5.5

claude-opus-5-5

Complex writing, coding agents, and knowledge work

Strength

Current Opus with adaptive reasoning and long context

Weakness

More expensive than Sonnet for routine work

Context and pricing notes

1M-token context; up to 128K output tokens.

How versions differ

$4 / $20 per million tokens, down from Opus 5 at $5 / $25.

OpenAI · version Sol

gpt-6-sol

Daily coding, tool use, and general reasoning

Strength

Balances capability and cost in the GPT-6 lineup

Weakness

Astra is the stronger option for the hardest tasks

Context and pricing notes

1.05M-token context; base rates shown, extra-long prompts may cost more.

How versions differ

$2 / $10 per million tokens; one fifth of Astra's base rates.

OpenAI · version Luna

gpt-6-luna

Focused tasks, simple edits, and high-volume processing

Strength

Lowest base rates in the GPT-6 lineup

Weakness

Validate quality before using it for complex reasoning

Context and pricing notes

1.05M-token context; up to 128K output tokens.

How versions differ

$0.10 / $0.50 per million tokens, below GPT-4o mini's base rates.

Google · version 3.8 Flash

gemini-3.8-flash

Long documents, coding, and multi-step tool use

Strength

Current stable Flash for agentic and everyday work

Weakness

Promotional rates expire December 31, 2026

Context and pricing notes

Long-context model. Output pricing includes thinking tokens.

How versions differ

$0.75 / $3.75 through 2026; $1.50 / $7.50 from January 1, 2027.

Google · version 3.5 Flash-Lite

gemini-3.5-flash-lite

High-volume extraction, translation, and simple processing

Strength

Low-cost stable Gemini for new projects

Weakness

Use Flash or Pro for harder reasoning

Context and pricing notes

Long-context processing; output pricing includes thinking tokens.

How versions differ

$0.30 / $2.50 per million tokens; a current alternative to 2.5 Flash.

DeepSeek · version V4.1 Flash

deepseek-flash

Budget coding, tool use, and visual inputs

Strength

Thinking and non-thinking modes with a 1M-token context

Weakness

Peak rates depend on time of day; test reliability on your tasks

Context and pricing notes

1M-token context; use the deepseek-flash API alias.

How versions differ

Peak uncached rates: $0.30 / $1.20. Off-peak rates are half.

DeepSeek · version V4 Pro 0813

deepseek-v4-pro

More demanding reasoning and coding in the DeepSeek lineup

Strength

Long context and thinking mode at moderate rates

Weakness

No vision support; more expensive than Flash

Context and pricing notes

1M-token context; current version is DeepSeek-V4-Pro-0813.

How versions differ

Peak uncached rates: $1.32 / $3.96. Off-peak rates are half.

Alibaba Cloud · version 3.8 Max

qwen3.8-max

Complex reasoning, coding, and long-document work

Strength

Current Qwen flagship with thinking, tools, and 1M context

Weakness

Higher cost than Plus or Flash; deployment region matters

Context and pricing notes

1M-token context. qwen3.8-max-0902 is also listed as a snapshot.

How versions differ

International rates: $2 / $6 per million tokens up to 1M input.

Alibaba Cloud · version 3.7 Plus

qwen3.7-plus

Balanced coding, office work, and tool-calling agents

Strength

Provider-recommended balance of capability and cost

Weakness

Higher price tier above 256K input; promotional discounts vary

Context and pricing notes

1M-token context; thinking and non-thinking output have the same listed rate.

How versions differ

International list rates: $0.40 / $1.60 up to 256K; $1.20 / $4.80 above. Promotions excluded.

Alibaba Cloud · version 3.8 Flash

qwen3.8-flash

High-volume processing, summaries, and everyday questions

Strength

Low international rates with 1M context and tool calling

Weakness

Compare quality with Plus or Max on difficult reasoning

Context and pricing notes

1M-token context. Confirm availability and pricing in your deployment region.

How versions differ

International rates: $0.15 / $0.47 per million tokens up to 1M input.

xAI · version 4.7

grok-4.7

General chat, coding, and tool-assisted research

Strength

Current xAI model for text and code with 500K context

Weakness

Live information requires search tools; long prompts cost more

Context and pricing notes

500K-token context. Higher-tier rates apply to the whole request.

How versions differ

$2 / $6 per million tokens below 200K input; $4 / $12 at or above 200K.

Moonshot AI · version K3

kimi-k3

Long-context reasoning and multi-step coding work

Strength

Current Kimi with a 1,048,576-token context window

Weakness

Cache writes are billed separately from base input and output

Context and pricing notes

About 1M context. API cache writes: $3/M for 5min or $6/M for 1h.

How versions differ

$3 / $15 per million tokens. Cached input is $0.30; cache-write rates depend on TTL.

Mistral AI · version 3.5

mistral-medium-3-5

Agentic coding, document questions, and multimodal work

Strength

Tool calling and structured outputs; open weights under Modified MIT

Weakness

Smaller context than 1M-token alternatives; review license terms

Context and pricing notes

256K-token context. Hosted API pricing is separate from self-hosting costs.

How versions differ

$1.50 / $7.50 per million tokens; Small 4 is the lower-cost Mistral option.

Mistral AI · version 4

mistral-small-2603

Budget instruction following, reasoning, and coding

Strength

Hybrid model with tool calling and Apache 2.0 open weights

Weakness

Test hard tasks against Medium before choosing solely on price

Context and pricing notes

256K-token context; dated API snapshot mistral-small-2603.

How versions differ

$0.15 / $0.60 per million tokens; much lower base rates than Medium 3.5.

Anthropic · version 5.1

claude-fable-5-1

Long unattended coding and knowledge-work agents

Strength

Built for long-horizon autonomous tasks in Copilot

Weakness

Expensive. Admins often must enable it. Data-retention rules differ from other Claude models.

Context and pricing notes

Meant for long agent runs. Don’t dump trivia into it.

How versions differ

Costs like a flagship ($10 / $50). Use for multi-step agents, not daily chat. Sonnet 5 is cheaper for normal coding.

Anthropic · version 5

claude-sonnet-5

Daily coding, writing, and planning at a lower Sonnet price

Strength

Newest Sonnet, cheaper than 4.5 / 4.6

Weakness

If a company still requires Sonnet 4.6, you cannot swap silently

Context and pricing notes

Excellent memory for most projects.

How versions differ

$2 to read / $10 to write per million tokens — less than Sonnet 4.6’s $3 / $15.

Anthropic · version 4.5

claude-haiku-4-5

Quick replies, light edits, cheap Claude volume

Strength

Fast and inexpensive, still a Claude

Weakness

Weaker on hard reasoning than Sonnet

Context and pricing notes

Fine for short threads. Don’t dump huge files.

How versions differ

Newer than retired Haiku 3.5. Slightly higher list price, stronger quality.

OpenAI · version Astra

gpt-6-astra

Long coding agents and hard multi-step work

Strength

Newest OpenAI flagship. Plans, checks, and finishes agent tasks with fewer loops

Weakness

Very expensive in Copilot credits. Not on Copilot Free/Pro in all rollouts

Context and pricing notes

Huge context, but long prompts cost extra after ~272k tokens.

How versions differ

$10 / $50 per million tokens at base rates. GPT-6 Sol costs $2 / $10; choose Astra when its extra capability justifies the cost.

Google · version 3.1 Pro

gemini-3.1-pro-preview

Long documents that still need careful answers

Strength

Newer Pro with huge context

Weakness

Overkill for short chats

Context and pricing notes

Base rates up to 200K input tokens; above that, $4 / $18 per million.

How versions differ

Use 3.8 Flash for simpler questions. Preview availability and limits can change.