Skip to content
ModelCap

Step 3.5 Flash

All-versions archive #173Arena #158Official publisher releaseOpen weights

StepFun·stepfun/step-3.5-flash

Step 3.5 Flash is ModelCap archive rank #173 (Measured), from the ModelCap Index as of 22 Sept 2026, 02:49 UTC.

Step 3.5 Flash listed OpenRouter input price is $0.10 per 1M tokens and listed output price is $0.30 per 1M tokens, from OpenRouter listed prices as of 22 Sept 2026, 02:49 UTC.

Step 3.5 Flash has a published context limit of 262K tokens in the ModelCap catalogue as of 22 Sept 2026, 02:49 UTC.

Step 3.5 Flash has 1 published OpenRouter serving endpoint, from OpenRouter as of 22 Sept 2026, 02:49 UTC.

Step 3.5 Flash weight access is open weights, from ModelCap classification of published repository metadata as of 22 Sept 2026, 02:49 UTC.

Snapshot facts · download the public dataset · methodology

ModelCap Index · All-versions archive
50.3#173
Measured · 65% support · score interval 45.5–55.1
Evidence as of 13 Sept 2026, 00:00 UTC
Market Gravity
44.0Decreased by 0.1
Secondary market signal

Superseded by Step 3.7 Flash. It no longer holds a current-board position; its all-versions archive rank is #173.

Input
$0.10
per 1M tokens
Output
$0.30
per 1M tokens
Context
262K
tokens
Providers
1
endpoint
Gateway spend
—
latest · Vercel share

Providers

Uptime measured over the last 30 minutes
ProviderInputOutputContextQuantUptime
SiliconFlowfp8$0.10$0.30262Kfp8100.0%

Hugging Face Inference Providers

Exact repository identity

Provider economics and telemetry published by the Hugging Face Router. These are operational signals, not benchmark results.

ProviderInputOutputContextTTFTThroughputFeatures
deepinfra
live
$0.10$0.30262K477 ms110.4 tok/stools
featherless-ai
live
——————

Reported evaluation leads (10)

Not independent measurement

Structured rows reported in this exact model repository at the pinned revision below. They are the publisher talking about itself: never measured evidence, never a score input on their own. They remain quarantined until dataset revision, harness, configuration and identity pass ModelCap’s independent admission review.

Dataset / taskMetricValueRevisionProvenance
cais/hle
hle
—23.1
Unpinned
not published
Reported
Quarantined
harborframework/terminal-bench-2.0
terminalbench_2
—51
Unpinned
not published
Reported
Quarantined
Idavidrein/gpqa
diamond
—83.5
Unpinned
not published
Reported
Quarantined
internlm/WildClawBench
overall
—26.7
Unpinned
not published
Reported
Quarantined
internlm/WildClawBench
avg_time
—430
Unpinned
not published
Reported
Quarantined
internlm/WildClawBench
avg_cost
—6.6
Unpinned
not published
Reported
Quarantined
MathArena/aime_2026
MathArena/aime_2026
—96.67
Unpinned
not published
Reported
Quarantined
MathArena/hmmt_feb_2026
MathArena/hmmt_feb_2026
—86.36
Unpinned
not published
Reported
Quarantined
SWE-bench/SWE-bench_Verified
swe_bench_%_resolved
—74.4
Unpinned
not published
Reported
Quarantined
TIGER-Lab/MMLU-Pro
mmlu_pro
—84.4
Unpinned
not published
Reported
Quarantined

Other versions (1)

ModelCap Index · All-versions archive

Methodology
50.3
Measured · 3 public benchmark observations across 3 boards
All-versions archive
#173
Public rank basis
Measured (measured)
Index support
64.8% · strong
Index interval
45.5–55.1
Rank posterior
Not available
Top 5 / Top 10 probability
Not available / Not available
Posterior as of
Snapshot timestamp unavailable
Evidence as of
13 Sept 2026, 00:00 UTC
Identity binding
Exact catalogue product · stepfun/stepfun/step-3.5-flash · aggregates disclosed benchmark configurations · endpoint configuration not claimed · artifact metadata stepfun-ai/Step-3.5-Flash@ab446a3de5e171ea341227e24bb1f090e1b771f7 (not the evaluation revision)
Index observations used
3
Qualified Index evidence point
50.3
Bayesian Evidence Score · measured evidence input only

BES normalizes admitted public benchmark evidence for measured Index rows. Its legacy score and observables prior do not define the public language rank.

Observed capability
50.3
Evidence status
confirmed · 88% mass
General preference
51.4
Coding
52.4
Agents & tools
26.3
Reasoning
—
Evidence breadth
97%
Evidence coverage
65%
Market and catalogue signals · secondary, never the language Index rank
Usage (OpenRouter popularity)
7.2
Liquidity (providers × uptime)
98.8
Open reach (HF downloads)
88.8
Surface (context / tools / modalities)
75.3
Freshness
40.8
  • · OpenRouter popularity #213
  • · 1 live provider
  • · HF 30d downloads 113152
  • · Catalogue: context 262144, reasoning, tools
  • · First seen 2026-01-29

These adoption and deployment observations remain context only. They do not change this model's ModelCap Index score or rank.

BES evidence selected · measured input only

The bounded 0–100 capability score blends each source’s competitive placement with its published achievement, then combines capability families. Evidence breadth adds a modest uncertainty adjustment; price and popularity are not part of benchmark-led rank.

Deployment readiness

Metadata index
92.1
strong · 98% metadata coverage
Not capability
Artifact reproducibility
100.0
Access & legal clarity
100.0
Deployability
89.0
Evaluation provenance
77.0

Missing: evaluationProvenance.evaluation-dataset-revisions.

A metadata completeness and deployability index—not a safety certification, quality grade, or production-readiness claim.

Market activity

Venue coverage
Market Gravity
44.0
OpenRouter weekly popularity
#213 of 322

Market Gravity uses OpenRouter’s full-catalogue popularity order. Sparse Vercel top-ten observations appear only when published and do not affect the score.

Arena preference

Status
Ranked
Source
LMArena
Official rank
#158 of 402
Rating
1394.1
95% confidence interval
1390–1398
Votes
57K
Source published at
13 Sept 2026, 00:00 UTC
Category
overall
Identity match confidence
100%

Market Gravity breakdown

Usage55%
7.2

OpenRouter weekly popularity across the full model catalogue.

Liquidity25%
98.8

Independent providers versus the model's open or closed cohort, weighted by uptime.

Open reach15%
88.8

Hugging Face 30-day downloads for open models; neutral for closed models.

Freshness5%
40.8

Time since first public availability, on a six-month half-life.

Specification

Context window
262K tokens
Max output
66K tokens
Inputs
Text
Outputs
Text
Tokenizer
Other
Cached input tokens
Not offered
First seen on OpenRouter
29 Jan 2026

Capabilities

  • Supported: Reasoning
  • Supported: Tool use
  • Not supported: Structured output
  • Not supported: Response format
  • Not supported: Moderated

Weights & access

Open weights. Downloadable weights are published under an OSI or free-culture license in the current repository metadata. This describes weight availability, not whether the full training stack qualifies as Open Source AI.

Access
Open weights
Licence
apache-2.0
Pinned revision
ab446a3de5e1
Parameters
199.4B
Architecture
Step3p5ForCausalLM
Model type
step3p5
Hugging Face downloads (30d)
113K
Downloads (all time)
1.6M
Likes
836
Repository
stepfun-ai/Step-3.5-Flash
Repository updated
17 Mar 2026, 05:54 UTC
Base lineage
stepfun-ai/step-3.5-flash

Benchmark leaderboards featuring Step 3.5 Flash

All leaderboards
Public benchmark boards on which Step 3.5 Flash has a published result
LeaderboardScoreSource rankConfigurationPublished
Arena coding leaderboardArena (LMArena)14501444–1456#141 of 397Default13 Sept 2026
τ²-Bench Telecom (Artificial Analysis run)Artificial Analysis87.4%#82 of 436tau2:telecom:dual-control:pass-at-1:3-repeats:source-model="Step 3.5 Flash 2603":reasoning=true9 Sept 2026
Humanity's Last Exam (Artificial Analysis run)Artificial Analysis24.5%#151 of 615hle:may-2025:text-only-2158:no-tools:pass-at-1:source-model="Step 3.5 Flash 2603":reasoning=true9 Sept 2026
WildClawBench OpenClaw leaderboardWildClawBench26.7#33 of 34Default20 Jul 2026

Step 3.5 Flash: common questions

How does Step 3.5 Flash rank among AI models?

Step 3.5 Flash is a superseded version and is not on the current board as of 22 September 2026. Its archive position, where one exists, is shown on the page; the newer version carries the current rank.

How much does Step 3.5 Flash cost per 1M tokens?

Step 3.5 Flash is listed at $0.10 per 1M input tokens and $0.30 per 1M output tokens as of 22 September 2026. Prices come from the OpenRouter catalogue and re-observe on every refresh. For 1,000 requests using 2,000 input and 500 output tokens each, the listed rates imply $0.35 total (2M input + 0.5M output). This excludes caching, batch discounts, prompt-length tiers, tool charges and retries; verify the chosen endpoint's rate and limits before budgeting.

What is the context window of Step 3.5 Flash?

Step 3.5 Flash has a published context window of 262K tokens (262,144) and a maximum output of 66K tokens. Individual providers can serve less than the published maximum; the providers table lists each endpoint's own limit.

Is Step 3.5 Flash open-weight?

Open weights (apache-2.0). Downloadable weights are published under an OSI or free-culture license in the current repository metadata. This describes weight availability, not whether the full training stack qualifies as Open Source AI. The repository is stepfun-ai/Step-3.5-Flash on Hugging Face.

Which API providers serve Step 3.5 Flash?

One provider serves Step 3.5 Flash through OpenRouter as of 22 September 2026: SiliconFlow. Each provider's price, context limit, quantization and measured uptime are in the providers table above.

Which benchmarks has Step 3.5 Flash been evaluated on?

Step 3.5 Flash has published results on 4 public boards tracked by ModelCap as of 22 September 2026: Arena coding 1450 (#141 of 397), AA τ²-Bench Telecom 87.4% (#82 of 436), AA Humanity's Last Exam 24.5% (#151 of 615) and WildClawBench OpenClaw 26.7 (#33 of 34). Each board page ranks every tracked model on that benchmark; the ModelCap Index combines them with published uncertainty.

When was Step 3.5 Flash released?

Step 3.5 Flash first appeared in the catalogue on 29 Jan 2026. It has since been superseded by Step 3.7 Flash. Rank and price movements since then are recorded on the site's changes feed.