Laguna XS 2.1 vs Step 3.7 Flash

Poolside · US  |  StepFun · China · Updated June 2026

Quick verdict

Pick Laguna XS 2.1 for remarkable efficiency — 70.9% on swe-bench verified from only 3b active parameters or open weights under openmdw-1.1, shipped day one in bf16, fp8, nvfp4 and int4 across every major runtime. Pick Step 3.7 Flash for a 198b-parameter sparse moe vision-language model (11b active) built specifically for agentic coding and search workflows or scmp reported it 'outshines larger rivals' from deepseek and moonshot on some benchmarks despite its smaller active size. On a tight budget at scale, Laguna XS 2.1 is the value pick.

Laguna XS 2.1 (Poolside, US) and Step 3.7 Flash (StepFun, China) line up two different AI ecosystems against each other — a comparison that is as much about cost philosophy and openness as raw capability. Laguna XS 2.1 is a 33B open-weight coding MoE running on 3B active parameters — 70.9% SWE-Bench Verified and very cheap, but unproven. Step 3.7 Flash is stepFun's 198B MoE vision-language model built for coding and search agents - a distinct Chinese lab with genuinely competitive open weights. Their biggest split is price, and the breakdown below shows exactly how that plays out for your workload.

Key differences at a glance

Side-by-side specs

SpecLaguna XS 2.1Step 3.7 Flash
ProviderPoolside (US) StepFun (China)
ReleasedJuly 2, 2026 May 29, 2026
Context window256K (~393 pages) 256K (~393 pages)
Price (in/out)$0.1/$0.2 per 1M tokens $0.2/$1.15 per 1M tokens
Open weight?Yes — self-hostable Yes — self-hostable
Modalitiestext, code text, image, code
SWE-Bench Verified70.9% Not published
MRCR v2 @ 1MNot published Not published

Who wins what

Remarkable efficiency — 70.9% on SWE-Bench Verified from only 3B active parameters

Laguna XS 2.1

Step 3.7 Flash is comparatively weak here — benchmark 'outshines rivals' claims are from press coverage of specific tests, not a full independent leaderboard sweep

Open weights under OpenMDW-1.1, shipped day one in BF16, FP8, NVFP4 and INT4 across every major runtime

Laguna XS 2.1

A 33B open-weight coding MoE running on 3B active parameters — 70.9% SWE-Bench Verified and very cheap, but unproven — and it runs cheaper at $0.1/$0.2 per 1M tokens.

Cheap even on the paid tier, at roughly a sixth of GLM 4.7's input price

Laguna XS 2.1

At $0.1/$0.2 per 1M tokens it undercuts Step 3.7 Flash ($0.2/$1.15 per 1M tokens), and that gap compounds at volume.

A 198B-parameter sparse MoE vision-language model (11B active) built specifically for agentic coding and search workflows

Step 3.7 Flash

Laguna XS 2.1 is comparatively weak here — weak on harder agentic work (37.5 on Terminal-Bench 2.0), and its gain over XS.2 is barely above noise

SCMP reported it 'outshines larger rivals' from DeepSeek and Moonshot on some benchmarks despite its smaller active size

Step 3.7 Flash

Step 3.7 Flash lists sCMP reported it 'outshines larger rivals' from DeepSeek and Moonshot on some benchmarks despite its smaller active size among its strengths; Laguna XS 2.1 does not.

Open weights (Apache 2.0) at a low per-token price

Step 3.7 Flash

Step 3.7 Flash lists open weights (Apache 2.0) at a low per-token price among its strengths; Laguna XS 2.1 does not.

Lowest cost at scale

Laguna XS 2.1

At $0.1/$0.2 per 1M tokens, it is the cheaper of the two — the gap dominates the bill on high-volume workloads.

Which should you pick?

A cost-sensitive startup shipping high volume

Laguna XS 2.1

At $0.1/$0.2 per 1M tokens it undercuts Step 3.7 Flash, and on millions of tokens that margin decides the monthly bill.

Anyone whose priority is remarkable efficiency — 70.9% on swe-bench verified from only 3b active parameters

Laguna XS 2.1

It is specifically built for that.

Anyone whose priority is a 198b-parameter sparse moe vision-language model (11b active) built specifically for agentic coding and search workflows

Step 3.7 Flash

That is its strongest area.

An enterprise with regional data-residency rules

Laguna XS 2.1 or Step 3.7 Flash

Origin (US vs China) affects where data is processed and which compliance regime applies — check the provider's terms for your region.

Laguna XS 2.1: where it fits

A 33B open-weight coding MoE running on 3B active parameters — 70.9% SWE-Bench Verified and very cheap, but unproven. Released July 2, 2026 by Poolside, it is built for remarkable efficiency — 70.9% on SWE-Bench Verified from only 3B active parameters, open weights under OpenMDW-1.1, shipped day one in BF16, FP8, NVFP4 and INT4 across every major runtime, cheap even on the paid tier, at roughly a sixth of GLM 4.7's input price, and unusually transparent evaluation — it publishes its harness, step limits, and sandbox specs.

Its trade-offs are real: weeks old with no independent replication; every published score traces back to Poolside's own harness, the free endpoint trains on your inputs and outputs — disqualifying for proprietary code, which is its main use case, and weak on harder agentic work (37.5 on Terminal-Bench 2.0), and its gain over XS.2 is barely above noise. At $0.1 in / $0.2 out per million tokens, it sits in the budget price band.

Step 3.7 Flash: where it fits

StepFun's 198B MoE vision-language model built for coding and search agents - a distinct Chinese lab with genuinely competitive open weights. Released May 29, 2026 by StepFun, it is built for a 198B-parameter sparse MoE vision-language model (11B active) built specifically for agentic coding and search workflows, sCMP reported it 'outshines larger rivals' from DeepSeek and Moonshot on some benchmarks despite its smaller active size, open weights (Apache 2.0) at a low per-token price, and includes a 1.8B vision encoder for image understanding alongside text.

Its trade-offs: stepFun is a newer, less established lab than DeepSeek, Alibaba, or Moonshot, benchmark 'outshines rivals' claims are from press coverage of specific tests, not a full independent leaderboard sweep, and smaller ecosystem and less third-party documentation than the more established Chinese labs. At $0.2 in / $1.15 out per million tokens, it sits in the budget price band.

The bottom line for this matchup

This is less "which is smarter" and more "which ecosystem fits." Laguna XS 2.1 (US) and Step 3.7 Flash (China) differ on pricing philosophy, data-residency, and tooling as much as on raw scores. Laguna XS 2.1 is the cheaper option, which matters at volume. The pragmatic move is to run one real task through both and judge the outputs against your own constraints — including where your data is allowed to be processed.

Want both Laguna XS 2.1 and Step 3.7 Flash without two subscriptions? LumiChats gives you these plus 40+ models under one ₹69/day pass (about $1/day) — draft with one, cross-check with the other.

See pricing

Frequently asked questions

Is Laguna XS 2.1 or Step 3.7 Flash better for coding?

Public SWE-Bench figures are not available for Step 3.7 Flash, so the honest test is your own repository — run an identical real bug through both. By design, Laguna XS 2.1 leans toward remarkable efficiency — 70.9% on swe-bench verified from only 3b active parameters while Step 3.7 Flash leans toward a 198b-parameter sparse moe vision-language model (11b active) built specifically for agentic coding and search workflows, and that positioning usually predicts which feels better on your codebase.

Which is cheaper, Laguna XS 2.1 or Step 3.7 Flash?

Laguna XS 2.1 is cheaper — $0.1/$0.2 per 1M tokens vs $0.2/$1.15 per 1M tokens, roughly 2× apart on input.

Which has the bigger context window?

Both advertise 256K (~393 pages). Remember advertised ≠ usable: recall typically degrades before the ceiling.

Can I use both Laguna XS 2.1 and Step 3.7 Flash together?

Yes — a multi-model platform like LumiChats gives you Laguna XS 2.1, Step 3.7 Flash and 40+ others under one ₹69/day pass (about $1/day), so you can draft with one and cross-check with the other instead of buying two subscriptions.

Which is newer, Laguna XS 2.1 or Step 3.7 Flash?

Laguna XS 2.1 — released July 2, 2026, about 34 days after Step 3.7 Flash.

Related comparisons

Specifications and benchmarks reflect publicly reported figures as of June 2026 and may change as providers release updates. Always verify on your own workload.