The Price War Claude Dethroned as DeepSeek and Xiaomi Dominate Coding LLMs

OpenRouter Ranking Shift: Claude Slips Out of Top 5 Coding Models as Price War Drives Agentic Adoption

The competitive landscape of Large Language Models (LLMs) for developers has witnessed a historic shift. According to the latest weekly platform statistics from OpenRouter, Anthropic’s Claude has slipped out of the top five most-used models in the programming category for the first time. Despite this drop, Claude remains a significant player in development workflows, with Claude Opus 4.7 holding the 8th position, capturing a 4% market share with a total volume of 129 billion tokens.

The latest weekly charts reveal that the top five most popular models in the programming category are now completely dominated by cost-efficient alternatives:

1. MiMo-V2.5

2. MiMo-V2.5-Pro

3. DeepSeek V4 Flash

4. DeepSeek V4 Pro

5. MiniMax M3

This dramatic reshuffle directly reflects recent aggressive price reductions by DeepSeek and Xiaomi's MiMo. Both providers have slashed rates for their respective Pro models to just $0.435 per million input tokens and $0.87 per million output tokens, while driving prompt caching costs down to an incredibly low $0.0036 per million tokens. These historic price cuts drastically lower the financial barrier for complex agentic workflows such as autonomous AI coding assistants that require continuous interaction with massive codebases. Meanwhile, MiniMax M3 is also maintaining high volume due to its 50% introductory launch promotion, pricing its tokens at $0.30 input and $1.20 output per million, with a prompt cache fee of $0.06.

While Claude has lost its historical monopoly in the programming sector where Claude Opus and Sonnet previously commanded a combined 50% market share last year Anthropic still maintains a massive overall footprint. When analyzing total usage across all categories on OpenRouter, Claude Sonnet 4.6 securely holds its position as the #4 most popular model overall, while Claude Opus 4.7 ranks at #6. However, competition remains fierce as MiMo-V2.5 continues its rapid ascent, leapfrogging into the #3 spot for total platform-wide volume.

The lower price and very low prompt caching in the Pro model are a significant turning point for developers. Current AI agents don't simply communicate line-by-line; they read entire projects' code, execute repetitive commands, and constantly use tool calls. The price drop to less than $1 makes running automated coding bots truly affordable and accessible at the enterprise level.

While technical and development teams may opt for more cost-effective models like DeepSeek and MiMo for large-scale tasks, top brands like Anthropic (Claude) remain strong and popular for general tasks. This reflects a clear shift in OpenRouter users' model categorization based on cost-effectiveness and application needs.

 

Microsoft Reaches GA for Coreutils for Windows via Winget. 

 

Source: OpenRouter 

💬 AI Content Assistant

Ask me anything about this article. No data is stored for your question.

OpenRouter Ranking Shift: Claude Slips Out of Top 5 Coding Models as Price War Drives Agentic Adoption

The competitive landscape of Large Language Models (LLMs) for developers has witnessed a historic shift. According to the latest weekly platform statistics from OpenRouter, Anthropic’s Claude has slipped out of the top five most-used models in the programming category for the first time. Despite this drop, Claude remains a significant player in development workflows, with Claude Opus 4.7 holding the 8th position, capturing a 4% market share with a total volume of 129 billion tokens.

The latest weekly charts reveal that the top five most popular models in the programming category are now completely dominated by cost-efficient alternatives:

1. MiMo-V2.5

2. MiMo-V2.5-Pro

3. DeepSeek V4 Flash

4. DeepSeek V4 Pro

5. MiniMax M3

This dramatic reshuffle directly reflects recent aggressive price reductions by DeepSeek and Xiaomi's MiMo. Both providers have slashed rates for their respective Pro models to just $0.435 per million input tokens and $0.87 per million output tokens, while driving prompt caching costs down to an incredibly low $0.0036 per million tokens. These historic price cuts drastically lower the financial barrier for complex agentic workflows such as autonomous AI coding assistants that require continuous interaction with massive codebases. Meanwhile, MiniMax M3 is also maintaining high volume due to its 50% introductory launch promotion, pricing its tokens at $0.30 input and $1.20 output per million, with a prompt cache fee of $0.06.

While Claude has lost its historical monopoly in the programming sector where Claude Opus and Sonnet previously commanded a combined 50% market share last year Anthropic still maintains a massive overall footprint. When analyzing total usage across all categories on OpenRouter, Claude Sonnet 4.6 securely holds its position as the #4 most popular model overall, while Claude Opus 4.7 ranks at #6. However, competition remains fierce as MiMo-V2.5 continues its rapid ascent, leapfrogging into the #3 spot for total platform-wide volume.

The lower price and very low prompt caching in the Pro model are a significant turning point for developers. Current AI agents don't simply communicate line-by-line; they read entire projects' code, execute repetitive commands, and constantly use tool calls. The price drop to less than $1 makes running automated coding bots truly affordable and accessible at the enterprise level.

While technical and development teams may opt for more cost-effective models like DeepSeek and MiMo for large-scale tasks, top brands like Anthropic (Claude) remain strong and popular for general tasks. This reflects a clear shift in OpenRouter users' model categorization based on cost-effectiveness and application needs.

 

Microsoft Reaches GA for Coreutils for Windows via Winget. 

 

Source: OpenRouter