Sulat.com
AI models
$10 off the fastest DeepSeek V4.1 Flash, Kimi K3 and GLM 5.3 from Synthetic
EmpirioLabs AI logo

Model details

Step 3.5 Flash 2603

Step 3.5 Flash 2603 is a text-only language model released by StepFun, with third-party comparison gateways confirming the StepFun attribution rather than the reseller listed in catalog metadata. The model is profiled alongside reasoning-oriented peers on LLMBase, where it is presented as an entry in the "Flash" tier of StepFun's lineup. AIsa's model gateway similarly lists it as a text model under the StepFun provider namespace, indicating that the model is positioned for general text generation and conversational use rather than multimodal workflows. Its placement on aggregator pages suggests it is aimed at developers who want a lightweight StepFun alternative for everyday prompting and short-form reasoning tasks.

On the LLMBase comparison page, Step 3.5 Flash 2603 is scored at 19.5 on reasoning while being priced at a low input rate of $0.10 per million tokens, with the site framing it as a value-oriented option against stronger but more expensive reasoning competitors. The same comparison flags it as the strongest pick on blended value and price metrics in its head-to-head, while trailing larger reasoning-focused models on raw reasoning index and coding benchmarks. Practical fit, based on the evidence, is for cost-sensitive text workloads where moderate reasoning is acceptable and the developer prefers a small, inexpensive StepFun text model over heavier alternatives.

EmpirioLabs AIstep-3-5-flash-2603

Quick Info

Powered by
Provider
EmpirioLabs AI
Model key
step-3-5-flash-2603
Release date
Apr 2, 2026
Last updated
Apr 2, 2026
Knowledge cutoff
2025-01
Input modalities
Output modalities
Capabilities

Cost

Input token cost
$0.10
Output token cost
$0.30

Limits

Input tokens
256,000 tokens
Output tokens
131,072 tokens
Context window
256,000 tokens

Latest news about Step 3.5 Flash 2603

EmpirioLabs AI

CoverageBenchmark

Independent benchmarking aggregator Artificial Analysis explicitly names Step 3.5 Flash 2603 and lists it among StepFun's three tracked models. The reported metrics are: Artificial Analysis Intelligence Index score of 27, output speed of 206 tokens/second, time-to-first-answer-token of 10.74 seconds, blended price of $ The page situates Step 3.5 Flash 2603 as the cheapest and second-fastest option in StepFun's lineup while being mid-tier in capability, with the intelligence score composed of nine evaluations (GDPval-AA v2, τ³-Banking, Terminal-Bench v2.1, SciCode, Humanity's Last Exam, GPQA Diamond, CritPt, AA-Omniscience, AA-LCR). S

EmpirioLabs AI

CoverageDiscourse

A HuggingFace community discussion titled "Is Step 3.5 Flash 2603 coming to huggingface?" was opened by user jukofyork on April 6, 2026, and references a StepFun X/Twitter post (status/2039711817794994451) as the source announcement for the model. Maintainer ZeroAurora replied on May 13 confirming "sadly no," indicatin The thread documents developer demand for open weights and clarifies the current HuggingFace availability status for the 2603 variant specifically, distinguishing it from the base Step-3.5-Flash and the later Step-3.7-Flash. It functions as official-channel-adjacent evidence via the linked StepFun social media announce

Videos about Step 3.5 Flash 2603