The LLM benchmark aggregator.

Palmyra X

Writer

Strongest in Safe-prompt compliance (#7 of 82), weakest in Agents (#165 of 267). Above par in 4 of 6 scopes. Among the models it meets almost everywhere, it finishes behind Claude Sonnet 5 and GPT-5.6 Terra and ahead of Grok 4.1 Fast and MiniMax M3.

§ 1 · Profile

What it is good at

Bars run from 50 — the average of the models each source lists — so right of the line is above par. The middle column is the gap to whoever leads that scope.

Safety53.9−7.6#72/3361/3

Safe-prompt compliance59.5−0.9#7/821/1

Fairness55.8−14.2#62/2991/2

Harm refusal52.6−9#118/2991/2

Agents46−22.1#165/2671/5

Tool use42.8−31.2#55/811/1

§ 2 · Head to head

What it beats, and what beats it

The same models turn up scope after scope. Counted once: where both were placed, who finished higher. Bars run right for Palmyra X, left for the other.

OLMoE 1B-7B Instructwon 4 · lost 0

Mistral Instruct v0.1 (7B)won 4 · lost 0

DBRX Instructwon 4 · lost 0

OLMo 2 13B Instructwon 4 · lost 0

Mistral Instruct v0.3 (7B)won 4 · lost 0

Command Rwon 4 · lost 0

Mixtral (8x7B)won 4 · lost 0

OLMo 2 7B Instructwon 4 · lost 0

Grok 3won 4 · lost 0

GPT-3.5 Turbowon 4 · lost 0

Llama 3.1 Instruct Turbo (70B)won 4 · lost 0

Command R Pluswon 4 · lost 0

DeepSeek LLM 67B Chatwon 4 · lost 0

Marin 8B Instructwon 4 · lost 0

Mixtral Instruct (8x22B)won 4 · lost 0

Mistral Small 3won 4 · lost 0

Palmyra Medwon 4 · lost 0

Gemini 2.0 Flash Litewon 4 · lost 0

Llama 3.1 Instruct Turbo (8B)won 4 · lost 0

Llama 3.1 Instruct Turbo (405B)won 4 · lost 0

Qwen1.5 72B Chatwon 4 · lost 0

Gemini 1.5 Flashwon 4 · lost 0

Gemini 2.0 Flashwon 4 · lost 0

IBM Granite 4.0 Smallwon 4 · lost 0

Llama 3 70B Instructwon 4 · lost 0

Phi-4 Mini Instructwon 4 · lost 0

Gemma 4 E2Bwon 4 · lost 0

DeepSeek V3won 5 · lost 0

Llama 3.1 8B Instructwon 5 · lost 0

Gemma 4 E4Bwon 5 · lost 0

Gemma 3 4B Itwon 5 · lost 0

Granite 3.2 8B Instructwon 5 · lost 0

Llama 4 Scout Instructwon 5 · lost 1

Gemma 3 27B Itwon 4 · lost 1

Gemma 3 12B Itwon 4 · lost 1

Llama 3.2 1B Instructwon 4 · lost 1

DeepSeek R1won 4 · lost 1

GPT-4o Miniwon 4 · lost 1

Llama 3.2 3B Instructwon 4 · lost 1

Nova Microwon 4 · lost 1

Gemma 4 31Bwon 4 · lost 1

Nova Prowon 4 · lost 1

Qwen3 8Bwon 4 · lost 1

Granite 3.1 8B Instructwon 4 · lost 1

IBM Granite 4.0 Micro (with guardian)won 3 · lost 1

IBM Granite 4.0 Microwon 3 · lost 1

DeepSeek R1 (hide reasoning)won 3 · lost 1

IBM Granite 4.0 Small (with guardian)won 3 · lost 1

IBM Granite 3.3 8B Instructwon 3 · lost 1

Olmo 2 32B Instructwon 3 · lost 1

Gemini 1.5 Prowon 3 · lost 1

Llama 3 8B Instructwon 3 · lost 1

Qwen 2.5 Instruct Turbo (7B)won 3 · lost 1

Claude 3 Haikuwon 3 · lost 1

Gemini 2.0 Prowon 3 · lost 1

Grok 3 Miniwon 3 · lost 1

Palmyra X5won 3 · lost 1

GPT-4won 3 · lost 1

Gemma 4 12Bwon 3 · lost 1

NVIDIA Nemotron 3 Super 120B A12Bwon 3 · lost 1

Gemma 4 26B A4Bwon 3 · lost 1

Qwen3.5 35B A3Bwon 3 · lost 1

Ling 3.0 Flashwon 3 · lost 1

Grok 4.3won 3 · lost 1

Gemini 3.5 Flashwon 3 · lost 1

Nemotron 3 Ultra 550B A55Bwon 3 · lost 1

Kimi K2.7 Codewon 3 · lost 1

Mimo V2.5 Prowon 3 · lost 1

Grok Build 0.1won 3 · lost 1

Qwen3.7 Maxwon 3 · lost 1

Kimi K2.5won 3 · lost 1

Ling 3.0 Flash VLwon 3 · lost 1

Gemini 3.6 Flashwon 3 · lost 1

Kimi K2.6won 3 · lost 1

DeepSeek V4 Flashwon 3 · lost 1

DeepSeek V4 Prowon 3 · lost 1

GPT-6 Solwon 3 · lost 1

Mimo V2.6 Prowon 3 · lost 1

Qwen3.8 Maxwon 3 · lost 1

GPT-4.1 Nanowon 4 · lost 2

Llama 4 Maverick Instructwon 4 · lost 2

Grok 4won 4 · lost 2

O3 Miniwon 3 · lost 2

Gemini 2.5 Prowon 3 · lost 2

Qwen3 Next 80B A3B Thinkingwon 3 · lost 2

GPT OSS 20Bwon 3 · lost 2

Command R7Bwon 3 · lost 2

Mistral Smallwon 3 · lost 2

Qwen3 30B A3B Instructwon 3 · lost 2

Command Awon 3 · lost 2

DeepSeek V3.2won 3 · lost 2

GLM-5.2won 3 · lost 2

GLM-4.6won 3 · lost 2

MiniMax M3won 3 · lost 2

Grok 4.1 Fastwon 3 · lost 2

Qwen3 235B A22B Throughputwon 2 · lost 2

GLM-4.5 Airwon 2 · lost 2

Palmyra Finwon 2 · lost 2

GPT-4.5won 2 · lost 2

GPT-4owon 2 · lost 2

Qwen2 72B Instructwon 2 · lost 2

Claude 3 Sonnetwon 2 · lost 2

O1 Miniwon 2 · lost 2

GPT-4 Turbowon 2 · lost 2

O1won 2 · lost 2

Qwen3.5 122B A10Bwon 2 · lost 2

GLM-5.1won 2 · lost 2

Claude Sonnet 4.6won 2 · lost 2

Gemini 3.8 Flashwon 2 · lost 2

GPT-6 Lunawon 2 · lost 2

GPT-5.6 Solwon 2 · lost 2

GLM-5.3 Flashwon 2 · lost 2

Gemini 2.5 Flash Litewon 3 · lost 3

GPT-4.1 Miniwon 3 · lost 3

O4 Miniwon 3 · lost 3

Gemini 2.5 Flashwon 3 · lost 3

Kimi K2won 3 · lost 3

Gemini 3 Prowon 3 · lost 3

Claude 3.7 Sonnetwon 2 · lost 3

Mistral Largewon 2 · lost 3

GPT-5.6 Lunawon 2 · lost 3

O3won 2 · lost 4

GPT-4.1won 2 · lost 4

Claude Opus 4won 1 · lost 3

Qwen 2.5 Instruct Turbo (72B)won 1 · lost 3

Claude 3 Opuswon 1 · lost 3

Claude 3.5 Sonnetwon 1 · lost 3

Claude 3.5 Haikuwon 1 · lost 3

Qwen3.5 397B A17Bwon 1 · lost 3

GPT-5.4won 1 · lost 3

Claude Opus 4.7won 1 · lost 3

Phi 4won 1 · lost 4

GPT OSS 120Bwon 1 · lost 4

Claude Sonnet 4won 1 · lost 4

GPT-5.1won 1 · lost 4

GPT-5won 1 · lost 4

GPT-5 Miniwon 1 · lost 5

Claude Haiku 4.5won 1 · lost 5

GPT-5 Nanowon 1 · lost 5

Qwen3 235B A22B Instructwon 1 · lost 5

Claude Sonnet 4.5won 1 · lost 5

MiniMax M2.7won 0 · lost 4

Hy3won 0 · lost 4

Grok 4.5won 0 · lost 4

GPT-5.5won 0 · lost 4

Muse Spark 1.1won 0 · lost 4

Claude Opus 4.8won 0 · lost 4

Grok 4.6won 0 · lost 4

Grok 4.7won 0 · lost 4

Claude Opus 5won 0 · lost 4

Claude Opus 5.5won 0 · lost 4

Inklingwon 0 · lost 5

GPT-5.2won 0 · lost 5

GPT-5.6 Terrawon 0 · lost 5

Claude Sonnet 5won 0 · lost 5

§ 3 · Sources

Where the numbers come from

2 publications, 6 figures. Every one links to the page it was read from.

BFCL ↗1 measureread 2026-09-27

BFCL v4 27.87%

HELM Safety ↗5 measuresread 2026-09-27

HELM Safety · HarmBench 73.8%HELM Safety · SimpleSafetyTests 100%HELM Safety · Anthropic Red Team 98.6%HELM Safety · BBQ 95.5%HELM Safety · XSTest 98.4%

Badge

[![PublicAI Index](https://publicai.io/model-index/badge?model=palmyra-x)](https://publicai.io/model-index/m/palmyra-x)