Model

Inflection-1

CLOSED

Inflection AI · Inflection family · released Jun 22, 2023

Pi's original in-house model; GPT-3.5-class personal-AI chatbot tuned for empathetic conversation over coding or reasoning.

ReasoningCodingVisionFunction callingTool useAgentic

513.0

Elo · rank #300

Parameters

Undisclosed

Active params

—

Context

8K tokens

Architecture

Dense Transformer (architecture, dataset, and training procedure undisclosed)

License

Proprietary

Languages

—

API price (in/out)

No hosted API

Modalities

text

Benchmark results

Bar shows position within the tracked field; marker = field best

ARC-ChallengeReasoning94.2%#12

best: Llama 3.1 405B · 96.9%

ARC-EasyReasoning89.6%#7

best: Phi-3-medium (14B) · 97.7%

BIG-Bench HardReasoning69.9%#58

best: ERNIE 4.5 300B-A47B · 94.3%

GSM8KMath62.9%#110

best: Llama 3.1 405B · 96.8%

HellaSwagReasoning84.3%#50

best: Claude 3 Opus · 95.4%

HumanEvalCoding35.4%#141

best: Claude Opus 4.5 · 99.4%

LAMBADAReasoning78.5%#4

best: GPT-3 175B · 86.4%

MATH-500Math16.7%#184

best: GPT-5 · 99.4%

MBPPCoding43.8%#80

best: Llama-3.3-Nemotron-Super-49B v1 (Reasoning On) · 91.3%

MMLUKnowledge72.7%#133

best: OpenAI o3 · 92.9%

PIQAReasoning84.2%#11

best: GPT-4o mini · 93.1%

TriviaQAKnowledge70.3%#24

best: Sarvam-1 (2B) · 90.6%

WinoGrandeReasoning83.3%#17

best: PaLM 2 · 90.9%

Run it locally

Closed weights — available via API only. No local deployment.

Input / M tok

—

Output / M tok

—

Inflection family

Elo progression across releases

Inflection-1Jun 2023513.0 Inflection-2Nov 2023646.1 Inflection-2.5Mar 2024870.8

Resources

DOCSinflection.ai/inflection-1

API price — · each benchmark row carries its own source badge (see methodology)