StepFun Step 5 Preview review: is the cheap 600B model worth it?

The one-minute verdict

I have spent the last few years putting AI on live support queues, so I read every model launch the same way: not "is the benchmark pretty," but "could I hand this a real job tomorrow." On that test, Step 5 lands as a strong preview and a weak purchase, and the gap between those two is the whole review.

Here is the scorecard before the detail.

If you only take one line from this: the price story is legitimate, the benchmark story needs independent runs, and "cheap model" and "finished product" are not the same purchase.

What StepFun Step 5 Preview actually is

StepFun (阶跃星辰) is a Shanghai lab that has shipped the "Step" family across text, vision, and audio for a couple of years, with a healthy open-weights footprint on Hugging Face. Step 5 Preview is its new flagship, described on the product page as a frontier model "for complex work and agentic tasks," leaning hard toward software engineering and finance.

The framing StepFun keeps returning to is "advancing the Pareto frontier," the line of best possible intelligence-for-your-money trade-offs. Its own launch chart says it plainly: match the cheaper open tier on intelligence, undercut it on cost.

Note the claim baked into that chart: a 65% lower cost than GLM-5.3 and Kimi K3 at a comparable intelligence level. It is a vendor chart, so I read it as the pitch rather than the verdict, but the pitch is at least a coherent one.

The architecture: 600B total, 27B active

The headline spec everyone latched onto is the shape of the model: 600B total parameters, 27B active per token, per the product page. That is a sparse mixture-of-experts design, and it is the whole cost story in one number.

In a dense model, every parameter runs for every token. In a mixture-of-experts model, a small router picks a handful of "expert" sub-networks per token, so only a slice of the model fires. Step 5 holds a frontier-sized 600B in memory but only lights up 27B at a time, which is why StepFun can price it like a mid-tier model.

Not everyone buys that the math nets out in Step 5's favor. The active-parameter count came up fast on Hacker News:

"600b-a27b doesn't sound enticing. Also with the higher number of active parameters compared to GLM 5.3 flash and DeepSeek V4/4.1 flash, I don't see how they want to be more efficient at inference."

Tepix, Hacker News

That is a fair challenge. 27B active is heavier than some rival "flash" tiers, so the efficiency win is really about the intelligence you get per active parameter, not the active count on its own. It is exactly the kind of thing independent inference benchmarks will settle, and StepFun's own charts will not.

It is a reasoning model, with a reasoning_effort control set to low, medium, or high. Higher effort means a longer internal thinking trace, and because output tokens are billed inclusive of that trace, effort is a direct cost dial. Hold that thought, it matters for the pricing section.

Pricing: the part I actually liked

Pricing is the easiest thing in this launch to like. From the StepFun pricing page, per 1M tokens:

The 95% cache-hit discount on input is the standout, and it is a big deal for agentic workloads that re-send a large, mostly-static system prompt on every step. Output at $2.70 covers both the reasoning trace and the final answer, so a chatty high-effort run costs real money.

Because that output-plus-reasoning cost is easy to underestimate, here is a quick estimator you can plug your own numbers into.

There is also a free V0 tier for kicking the tires (5 concurrency, 10 requests per minute), with rate limits scaling by how much you have topped up. For raw price-per-token, this is one of the better deals in the reasoning-model class right now. It plays in the same budget bracket as DeepSeek V4 Flash rather than the premium tier, and undercuts most of it.

Benchmarks, and the caveat that dominates this review

StepFun published a full comparison against open peers (Kimi K3, GLM-5.3) and the closed frontier (GPT-6 Astra, Claude Opus 5). Here is the shape of it, all numbers from StepFun's own product page.

The read is consistent: Step 5 sits at parity with the other leading open models and behind the top closed models on raw coding quality, while winning on cost. It trails Claude Opus 5 most on the hard coding tests. Finance is its standout, where it edges the open peers and beats GPT-6 Astra. Its own StepCodeBench radar tells the same story against the open tier.

Now the caveat, and it is the reason my benchmark-credibility score is a 4. Every number above is self-reported, there is no technical report or paper yet, and StepCodeBench is StepFun's own benchmark (553 repositories, 33 languages). A model tends to look good on the test its own team designed. The community flagged the softer spots fast, especially verbosity:

"Based on the examples, it's like the thing took a writing class from Claude, but isn't quite as smart, kind of worst of both worlds. The 'Interactive Reporting' one is particularly terrible. A huge amount of waffly padding around a thesis that may or may not exist."

wrs, Hacker News

And one commenter caught the launch demo giving itself away, which is exactly the kind of thing that makes me want to run my own evals before trusting a launch chart:

"In their first demo video, to make a 3D render of the photo, the thinking trace gives away the game [...] I'm always astounded how little effort is put into checking the AI answers displayed in these announcements."

BoppreH, Hacker News

None of that makes Step 5 a bad model. It makes it an unproven one, which is a different thing. The correct response is to simulate it against your own workload rather than take the launch chart at face value, and that instinct carries straight over to how you should trust any AI in customer support.

Where it shines: long context and finance

Two strengths held up across everything I read. The 1M-token context is the spec I would lean on hardest. It is enough to drop an entire codebase, a quarter of support transcripts, or a stack of PDFs into a single call and ask for cross-document reasoning. On the multimodal side you can pass up to 60 images per request plus short video, though output is text only.

StepFun also leans into the coding story hard. It claims Step 5 can go "from software engineering and visual applications to programmable hardware," and that in expert evaluations about 70% of participants judged it capable of autonomously solving coding tasks of moderately high complexity, per its product page. There is a Claude Code integration through "Step Plan" that exposes the full 1M context, which is a smart place to meet developers. The finance result on FrontierFinance is the one benchmark where it clearly leads the pack, closed models included.

Pros and cons

Pulling the review together into the two lists I actually use when deciding whether to test something:

What is good

• Best price-for-intelligence in the cheaper tier, at $1.00 / $2.70 per 1M tokens.
• A real 1M-token context, not a marketing number capped at 128k in practice.
• 95% cache-hit discount that rewards agentic, repeated-prompt workloads.
• Honest positioning: it says "best of the cheaper tier," not "we beat the frontier."
• An open-weights release on the roadmap, with an Apache-2.0 precedent from the Flash line.

What gives me pause

• Every benchmark is self-reported, with no technical report and one home-made test.
• Community read is verbose output, which quietly inflates the output-token bill.
• 27B active is heavier than some rival flash tiers, so the efficiency win is unsettled.
• API-only today; the weights that make the "open" pitch interesting have not landed.
• It is a raw model, so the work of turning it into a product is entirely yours.

Who Step 5 is for, and who should wait

Here is where I land. If you are a developer or a lab that wants a capable, cheap, long-context model to build on, especially for coding, finance, or research over big document sets, Step 5 Preview is worth a real test. The price-per-intelligence story is legitimately good, and the open-weights promise makes it more interesting still.

If you are waiting for a settled, proven model, wait. It is a preview, the benchmarks are self-graded, and the community's early read is "promising but verbose and unproven." Even a friendly commenter put the pattern gently:

"The previous Step models were pretty decent, but unfortunately for them, their model reasons too much and too long [...] Hopefully this doesn't reason too long to get to the answer. I welcome any open model, the more the merrier."

segmondy, Hacker News

I will give StepFun credit for the one thing most launches get wrong, which is knowing what they are:

"Their posisitoning is nice. Instead of saying they are cheaper and a bit less performant (in terms of intelligence), they say they are best among the cheaper and a bit less performant ones."

ghoshbishakh, Hacker News

Where a raw model stops and a teammate starts

There is one framing worth pushing back on, and it is not StepFun's fault, it applies to every model launch. A model is not a solution. Step 5 Preview is an engine. It gives you an API, a token bill, and raw capability. Turning that into something that answers your customers means you build the app, connect your data, write the guardrails, and own the maintenance.

That gap is why eesel exists. Instead of a raw model, you hire a ready-to-work teammate for a specific job. The AI helpdesk teammate joins your existing queue, trains on your past tickets and help center, and drafts or sends replies inside the helpdesk you already run, whether that is Zendesk, Freshdesk, or Gorgias. Pricing is a flat rate per resolution with no per-seat fee, so cost maps to work done rather than tokens burned.

The part I care most about, given the self-reported-benchmark problem above, is that you can simulate the teammate on thousands of your own past tickets before it ever touches a live customer, so you see its real resolution rate on your workload instead of a vendor's chart. And if you would rather drive it from a terminal than a dashboard, eesel ships a real CLI and MCP server so coding agents like Claude Code can operate the same teammate directly. New models like Step 5 are exciting infrastructure. A teammate is what actually shows up for the shift. You can try eesel free.