ModelsCompareBest forBenchmarksStatusPricingAPI
Compare / head-to-head

GPT-5.4vsStep 3.5 Flash

GPT-5.4 leads 4 of 4 shared benchmarks. GPT-5.4 has the larger context window (1.1M tokens).

Benchmarks from cited public sources; pricing from official pages; status from official provider feeds.

Shared benchmarks
4 – 0
GPT-5.4 leads
Cheaper per token
list price, input + output
Larger context
GPT-5.4
1.1M tokens
Provider uptime (30d)
95.122% · —
OpenAI · StepFun
GPT-5.4
OpenAI · released 2026-03-05
textvision
Context
1.1M
Max out
128K
Input /1M
$2.5
Output /1M
$15
Cached /1M
$0.25
Scores
21 · 11 core
Step 3.5 Flash
StepFun · released 2026-02-12
text
Context
256K
Max out
Input /1M
Output /1M
Cached /1M
Scores
18 · 5 core
Quality

Benchmark matrix

BenchmarkGPT-5.4Step 3.5 FlashΔEdge
Reported by both · 4
AIME 202699.2% matharena96.7% matharena+2.5 ptGPT-5.4
LMArena Elo1452.6 1403.7 +48.9GPT-5.4
BrowseComp82.7% 51.6% +31.1 ptGPT-5.4
Terminal-Bench 2.075.1% 51% +24.1 ptGPT-5.4
Only GPT-5.4 reports · 16
GPQA Diamond92.8% not reported
Humanity's Last Exam (no tools)39.8% not reported
FrontierMath Tier 4 v2 (Epoch AI run)49% epochnot reported
FrontierMath Tiers 1-3 v2 (Epoch AI run)78.6% epochnot reported
GDPval (wins or ties)83% not reported
Humanity's Last Exam (with tools)52.1% not reported
MCP Atlas67.2% not reported
MMMU-Pro (no tools)81.2% not reported
MMMU-Pro (with tools)82.1% not reported
OSWorld-Verified75% not reported
OTIS Mock AIME 2024-2025 (Epoch AI run)97.8% epochnot reported
SimpleQA Verified45.1% epochnot reported
SWE-Bench Pro57.7% not reported
SWE-bench Verified (Epoch AI run)76.9% epochnot reported
Tau2-bench Telecom98.9% not reported
Toolathlon54.6% not reported
Only Step 3.5 Flash reports · 13
SWE-bench Verifiednot reported74.4%
AIME 2025not reported97.3%
BrowseComp (with context manager)not reported69%
BrowseComp-ZHnot reported66.9%
BrowseComp-ZH (with context manager)not reported73.7%
GAIA (no file)not reported84.5%
HMMT 2025 (Feb/Nov average)not reported96.2%
IMOAnswerBenchnot reported85.4%
LiveCodeBench v6not reported86.4%
ResearchRubricsnot reported65.3%
Tau2-Benchnot reported88.2%
xbench-DeepSearch 2025.05not reported83.7%
xbench-DeepSearch 2025.10not reported56.3%
Scores tagged "epoch" or "matharena" are independent runs, used only where the lab has not published its own; ⚠ marks rows MathArena flags as released after the competition. Higher is better on every row. Δ is GPT-5.4 minus Step 3.5 Flash in the benchmark's own unit. "Not reported" means the lab has not published that figure; it is not a zero. ↗ opens the source.
Specs & pricing

Side by side

SpecGPT-5.4Step 3.5 FlashEdge
Context window1.1M tokens256K tokensGPT-5.4
Max output128K tokens
Input price / 1M$2.5
Output price / 1M$15
Cached input / 1M$0.25
Input + output / 1M
Lower is cheaper. List prices; batch, tool and regional fees excluded.
$17.5
Modalitiestext · visiontextGPT-5.4
Released2026-03-052026-02-12
Cited benchmark scores2118
Reliability

Provider status

All providers
More matchups

Step 3.5 Flash vs …

Built by Respan
Which one wins on your data?

Public benchmarks are a starting point. Run GPT-5.4 and Step 3.5 Flash on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.