Skip to content

A reactive notebook for Python — run reproducible experiments, query with SQL, execute as a script, deploy as an app, and version with git. Stored as pure Python. All in a modern, AI-native editor.

feat: add new W&B models to llm-info data (#9465)

Myles Scolnicke06cc2feef53
diff --git a/packages/llm-info/data/models.yml b/packages/llm-info/data/models.yml
index 1366884fb4..fa89a962fd 100644
--- a/packages/llm-info/data/models.yml
+++ b/packages/llm-info/data/models.yml
@@ -226,6 +226,41 @@
# Weights & Biases Models
+- name: DeepSeek V3.1
+ model: deepseek-ai/DeepSeek-V3.1
+ description: Large hybrid Mixture-of-Experts model that supports both thinking and non-thinking modes
+ providers: [wandb]
+ roles: [chat, edit]
+ thinking: true
+
+- name: Google Gemma 4 31B
+ model: google/gemma-4-31B-it
+ description: Vision and text model designed for advanced reasoning, agentic workflows, and longer context
+ providers: [wandb]
+ roles: [chat, edit]
+ thinking: true
+
+- name: IBM Granite 4.1 8B
+ model: ibm-granite/granite-4.1-8b
+ description: Long-context text model with enhanced tool calling and instruction following
+ providers: [wandb]
+ roles: [chat, edit]
+ thinking: false
+
+- name: Meta Llama 3.3 70B (W&B)
+ model: meta-llama/Llama-3.3-70B-Instruct
+ description: Multilingual model excelling in conversational tasks, detailed instruction-following, and coding
+ providers: [wandb]
+ roles: [chat, edit]
+ thinking: false
+
+- name: Meta Llama 3.1 70B (W&B)
+ model: meta-llama/Llama-3.1-70B-Instruct
+ description: Multilingual chatbot model optimized for responsiveness
+ providers: [wandb]
+ roles: [chat, edit]
+ thinking: false
+
- name: Meta Llama 3.1 8B (W&B)
model: meta-llama/Llama-3.1-8B-Instruct
description: Efficient conversational model optimized for responsive multilingual chatbot interactions
@@ -233,6 +268,41 @@
roles: [chat, edit]
thinking: false
+- name: Microsoft Phi 4 Mini 3.8B
+ model: microsoft/Phi-4-mini-instruct
+ description: Compact instruction-tuned model for resource-constrained environments
+ providers: [wandb]
+ roles: [chat, edit]
+ thinking: false
+
+- name: MiniMax M2.5
+ model: MiniMaxAI/MiniMax-M2.5
+ description: Mixture-of-Experts model with a highly sparse architecture designed for high-throughput and low latency
+ providers: [wandb]
+ roles: [chat, edit]
+ thinking: true
+
+- name: Moonshot AI Kimi K2.6
+ model: moonshotai/Kimi-K2.6
+ description: Multimodal Mixture-of-Experts model with 32B activated parameters and 1T total parameters
+ providers: [wandb]
+ roles: [chat, edit]
+ thinking: true
+
+- name: Moonshot AI Kimi K2.5
+ model: moonshotai/Kimi-K2.5
+ description: Multimodal Mixture-of-Experts model optimized for complex tool use, reasoning, and code synthesis
+ providers: [wandb]
+ roles: [chat, edit]
+ thinking: true
+
+- name: NVIDIA Nemotron 3 Super 120B
+ model: nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8
+ description: LatentMoE model for agentic, reasoning, and conversational capabilities
+ providers: [wandb]
+ roles: [chat, edit]
+ thinking: true
+
- name: OpenAI GPT OSS 120B
model: openai/gpt-oss-120b
description: Efficient Mixture-of-Experts model designed for high-reasoning, agentic and general-purpose use cases
@@ -247,37 +317,51 @@
roles: [chat, edit]
thinking: true
-- name: DeepSeek V3-0324
- model: deepseek-ai/DeepSeek-V3-0324
- description: Robust Mixture-of-Experts model tailored for high-complexity language processing and comprehensive document analysis
+- name: OpenPipe Qwen3 14B Instruct
+ model: OpenPipe/Qwen3-14B-Instruct
+ description: Multilingual instruction-tuned model optimized for agents
providers: [wandb]
roles: [chat, edit]
thinking: false
-- name: Meta Llama 3.3 70B (W&B)
- model: meta-llama/Llama-3.3-70B-Instruct
- description: Multilingual model excelling in conversational tasks, detailed instruction-following, and coding
+- name: Qwen3.5 35B A3B
+ model: Qwen/Qwen3.5-35B-A3B
+ description: Multimodal Mixture-of-Experts model for chat, reasoning, and agentic tasks
providers: [wandb]
roles: [chat, edit]
- thinking: false
+ thinking: true
-- name: DeepSeek R1-0528
- model: deepseek-ai/DeepSeek-R1-0528
- description: Optimized for precise reasoning tasks including complex coding, math, and structured document analysis
+- name: Qwen3 235B A22B Thinking-2507
+ model: Qwen/Qwen3-235B-A22B-Thinking-2507
+ description: Mixture-of-Experts model with reasoning capabilities for structured reasoning, math, and long-form generation
providers: [wandb]
roles: [chat, edit]
thinking: true
-- name: MoonshotAI Kimi K2
- model: moonshotai/Kimi-K2-Instruct
- description: Mixture-of-Experts model optimized for complex tool use, reasoning, and code synthesis
+- name: Qwen3 235B A22B Instruct-2507
+ model: Qwen/Qwen3-235B-A22B-Instruct-2507
+ description: Multilingual Mixture-of-Experts instruction model optimized for logical reasoning
providers: [wandb]
roles: [chat, edit]
- thinking: true
+ thinking: false
+
+- name: Qwen3 30B A3B Instruct-2507
+ model: Qwen/Qwen3-30B-A3B-Instruct-2507
+ description: Mixture-of-Experts model with enhanced reasoning, coding, and long-context understanding
+ providers: [wandb]
+ roles: [chat, edit]
+ thinking: false
+
+- name: Qwen3 Coder 480B A35B
+ model: Qwen/Qwen3-Coder-480B-A35B-Instruct
+ description: Mixture-of-Experts model optimized for agentic coding tasks such as function calling and tool use
+ providers: [wandb]
+ roles: [chat, edit]
+ thinking: false
-- name: Z.AI GLM 4.5
- model: zai-org/GLM-4.5
- description: Mixture-of-Experts model with user-controllable thinking/non-thinking modes for strong reasoning, code generation, and agent alignment
+- name: Z.AI GLM 5.1
+ model: zai-org/GLM-5.1
+ description: Mixture-of-Experts model for long-horizon agentic engineering and advanced reasoning
providers: [wandb]
roles: [chat, edit]
thinking: true