Update the correct naming for llama 3 70b instruct models

[juliettech13]

We already had the meta-llama/llama-3.3-70b-instruct model integrated, but the naming was wrong because of the groq provider offering the versatile version.

This PR fixes it so we display both.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Preview	Comments	Updated (UTC)
helicone	Ready	Preview	Comment	Oct 20, 2025 6:56pm
helicone-bifrost	Error			Oct 20, 2025 6:56pm
helicone-eu	Error			Oct 20, 2025 6:56pm

💡 Enable Vercel Agent with $100 free credit for automated AI reviews

[claude]

Claude finished @juliettech13's task —— View job

---

PR Review Analysis

Reviewing the changes for llama 3 70b instruct model naming fixes...

Tasks

• Read and analyze changed files
• Review cost model changes
• Check snapshot test updates
• Assess code quality and potential issues
• Provide comprehensive feedback with confidence score

Analysis Complete

The PR correctly addresses the naming conflict issue with the Meta Llama 3.3 70B models. Here's my detailed review:

✅ Positive Changes

1. Proper Model Separation: The PR correctly separates the "versatile" and "instruct" variants of the llama-3.3-70b model, which were previously conflated.

2. Accurate Model Configurations:

   • llama-3.3-70b-versatile now properly maps to Groq's "llama-3.3-70b-versatile" model ID
   • llama-3.3-70b-instruct correctly uses "meta-llama/llama-3.3-70b-instruct" for other providers
   • Context lengths and token limits are properly differentiated (131K vs 128K context, 32K vs 16K output)

3. Consistent Provider Mapping: Each model variant now maps to appropriate providers based on their actual offerings.

4. Updated Test Snapshots: All snapshot tests are properly updated to reflect the new model structure.

🔍 Technical Review

File: models.ts

• Lines 37-47: New llama-3.3-70b-versatile model definition is well-structured
• Lines 48-58: Updated llama-3.3-70b-instruct with correct specs
• Both models have appropriate descriptions, context lengths, and creation dates

File: endpoints.ts

• Line 184: Correctly renamed endpoint key from llama-3.3-70b-instruct:groq to llama-3.3-70b-versatile:groq
• Provider model IDs are accurate and match the intended model variants
• Pricing and parameter configurations are maintained correctly

File: registrySnapshots.test.ts.snap

• Snapshot updates properly reflect the new model structure
• Total model count increased from 66 to 67 PTB-enabled models
• Provider mappings are correctly updated

🎯 Score: 9/10 - Excellent implementation with proper model separation

Suggestions Summary:

• ✅ No critical issues found - The implementation correctly addresses the naming conflict
• ✅ Code quality is high - Follows existing patterns and conventions
• ✅ Breaking changes are minimal - Only affects internal model mapping, not external APIs
• Minor consideration: The descriptions could be slightly more concise, but they're informative and accurate

This PR successfully resolves the naming confusion and properly displays both variants of the Llama 3.3 70B models. The implementation is clean, well-tested, and follows the established codebase patterns.

---

[greptile-apps]

Greptile Overview

Summary

This PR correctly differentiates between two distinct Llama 3.3 70B model variants that were previously conflated:

• llama-3.3-70b-versatile - Groq's optimized variant with 131K context and 32K max output tokens
• llama-3.3-70b-instruct - Meta's standard instruct model with 128K context and 16K max output tokens

The changes ensure accurate model identification across providers:

• Groq now correctly maps to the "versatile" variant
• Novita and OpenRouter correctly map to the standard "instruct" model

Test snapshots have been regenerated to reflect the model registry now tracking both variants separately.

Confidence Score: 5/5

• This PR is safe to merge with no issues found
• The changes are straightforward and correct - they properly distinguish between two model variants that have different specifications. All provider mappings are accurate, test snapshots are appropriately updated, and the implementation follows the existing codebase patterns
• No files require special attention

Important Files Changed

File Analysis

Filename	Score	Overview
packages/cost/models/authors/meta/llama/models.ts	4/5	Correctly splits llama-3.3-70b into two distinct models: versatile (131K context, 32K max output) and instruct (128K context, 16K max output) variants with appropriate descriptions
packages/cost/models/authors/meta/llama/endpoints.ts	5/5	Updates Groq endpoint key from llama-3.3-70b-instruct to llama-3.3-70b-versatile to match Groq's actual model identifier
packages/tests/cost/snapshots/registrySnapshots.test.ts.snap	5/5	Test snapshot correctly updated to reflect the model split - Groq now offers versatile variant, while Novita and OpenRouter offer the standard instruct model

Sequence Diagram

sequenceDiagram
    participant User
    participant Registry
    participant Groq
    participant Novita
    participant OpenRouter

    Note over Registry: Before: Single model entry<br/>"llama-3.3-70b-instruct"<br/>mapped to all providers

    User->>Registry: Request llama-3.3-70b model info
    Registry->>Groq: Returns "llama-3.3-70b-instruct"
    Note over Groq: ❌ Incorrect!<br/>Groq uses "versatile" variant

    Note over Registry: After: Two distinct model entries

    User->>Registry: Request llama-3.3-70b-versatile
    Registry->>Groq: Returns "llama-3.3-70b-versatile"
    Note over Groq: ✅ Correct!<br/>131K context, 32K output

    User->>Registry: Request llama-3.3-70b-instruct
    Registry->>Novita: Returns "llama-3.3-70b-instruct"
    Registry->>OpenRouter: Returns "llama-3.3-70b-instruct"
    Note over Novita,OpenRouter: ✅ Correct!<br/>128K context, 16K output

Loading

_{2 files reviewed, no comments}

_{Edit Code Review Agent Settings | Greptile}