MiniMax M2.7 vs Claude Opus 4.6: A Cost-Effective Coding Task Benchmark

MiniMax M2.7 vs Claude Opus 4.6: A Cost-Effective Coding Task Benchmark

MiniMax M2.7, launched on March 18, scored a respectable 56.22% on the SWE-Pro benchmark, which is close to Claude Opus 4.6's performance. To see how these models fare in real-world coding tasks, we ran both through three tests using Kilo Code, an AI-assisted coding environment for VS Code. The results are compelling, especially when it comes to cost-effectiveness.

Key Takeaways

• Performance: Both models found all 6 bugs and all 10 security vulnerabilities.
• Cost: MiniMax M2.7 is significantly cheaper at $0.30/$1.20 per million tokens (input/output) compared to Claude Opus 4.6’s $5/$25, a 17x difference on input and 21x on output.
• Quality: MiniMax M2.7 delivered 90% of the quality for just 7% of the cost.

Test Design

We created three TypeScript codebases to test both models in Kilo Code's Code mode, ensuring they received the same prompts without any hints. Each model was scored independently after all tests were completed.

• Test 1: Full-Stack Event Processing System (35 points)
  • Build a complete system from a spec, including an async pipeline, WebSocket streaming, and rate limiting.
• Test 2: Bug Investigation from Symptoms (30 points)
  • Trace 6 bugs from production log output to root causes and fix them.
• Test 3: Security Audit (35 points)
  • Find and fix 10 planted security vulnerabilities across a team collaboration API.

Test 1: Full-Stack Event Processing System

For this test, we gave both models the following prompt:

"Build a real-time event processing system in TypeScript from the specification in @SPEC.md. Use Hono for the web framework, Prisma with SQLite for the database, Zod for input validation, and ws for WebSocket support."

The spec required 7 components:

• Event ingestion API with API key auth
• Async processing pipeline with exponential backoff retry
• Event storage with processing history
• Query API with pagination and filtering
• WebSocket endpoint for live streaming
• Per-key rate limiting
• Health/metrics endpoints

Both models successfully implemented all 7 components. However, the score difference came from code organization and test coverage.

Architecture Details

• Claude Opus 4.6:

  • More thorough fixes
  • 2x more tests
  • Better code organization

• MiniMax M2.7:

  • Efficient implementation of all components
  • Slightly less comprehensive testing and documentation

Test 2: Bug Investigation from Symptoms

In this test, both models were tasked with tracing 6 bugs from production log output to their root causes and fixing them. Both models identified and fixed all 6 bugs, but Claude Opus 4.6 provided more detailed explanations and better-documented fixes.

Test 3: Security Audit

For the security audit, both models had to find and fix 10 planted security vulnerabilities across a team collaboration API. Both models successfully found and fixed all 10 vulnerabilities, with Claude Opus 4.6 again providing more thorough documentation and additional test cases.

Cost Analysis

The cost difference between MiniMax M2.7 and Claude Opus 4.6 is substantial:

• MiniMax M2.7: $0.30 per million tokens (input), $1.20 per million tokens (output)
• Claude Opus 4.6: $5 per million tokens (input), $25 per million tokens (output)

For the coding tasks, MiniMax M2.7 cost a total of $0.27, while Claude Opus 4.6 cost $3.67. This represents a significant cost savings without a substantial loss in quality.

Conclusion

Both MiniMax M2.7 and Claude Opus 4.6 performed well in our