NotAnotherBench

GROUNDED IN REAL HUMAN CONVERSATIONSYOUR BENCH

LOADING DATASHEET

Leaderboard Writing Coding Agents Research This week Launches Best for Labs Compare Stats Methodology

made by ani.computer

NotAnotherBench

GROUNDED IN REAL HUMAN CONVERSATIONSYOUR BENCH

ALTERNATIVES

Phi 4 Mini alternatives

Models people actually talk about as stand-ins, plus same-category crowd favorites. More from Microsoft · Problems

12 PICKS

LFM2.5 2.6B6.4
Liquid · Text
Pick it for an edge-native model with native tool calling and a 128K context window
🪶 RUNS ON LESS🔒 PRIVATE & OFFLINE⚡ FAST COMPARE
Nemotron 3.5 Lightning6.3
Nvidia · Multimodal
Pick it for fast, agentic workflows and coding with only 3B active parameters
💻 STRONG AT CODE⚡ FAST🏆 NEAR-PEER QUALITY COMPARE
Phi 45.5
Microsoft · Multimodal
Pick it for higher quality reasoning and math if your hardware can handle 14B parameters
🏆 NEAR-PEER QUALITY💻 STRONG AT CODE COMPARE
DeepSeek V4 Flash7.2
DeepSeek · Text
People love: People feel it is worth the money
COMPARE
Ling 3.1 Flash7.2
inclusionAI · Text
People love: People feel it is worth the money
COMPARE
DeepSeek Reasoner6.8
DeepSeek · Text
People love: People feel it is worth the money
COMPARE
Claude Opus 5.56.7
Anthropic · Text
People love: People feel it is worth the money
COMPARE
Gemma 4 26B A4B6.7
Google · Text
People love: Popular for coding help
COMPARE
Claude Sonnet 56.6
Anthropic · Text
People love: Good at writing and wording things
COMPARE
R16.6
DeepSeek · Text
People love: People feel it is worth the money
COMPARE
Qwen3 0.6B6.6
Alibaba · Text
People love: Popular for coding help
COMPARE
Step 16.5
StepFun · Text
People love: People feel it is worth the money
COMPARE