Head-to-head AI model comparison graphic: Mistral Large 4 Preview (Mistral) vs Step 5 Preview (StepFun). Intelligence index 25.3 vs 28.5; output speed 130 vs 151 tokens per second; time to first token 1.9s vs 0.91s; output price $2.10 vs $2.70 per million tokens; input price $0.70 vs $1.00 per million tokens. franklineh.com
Mistral Large 4 Preview vs Step 5 Preview: head-to-head AI benchmark comparison. Step 5 Preview leads on intelligence index, output speed and time to first token; Mistral Large 4 Preview leads on output price and input price.
Head to headUpdated

Mistral Large 4 Preview vs Step 5 Preview compared

Step 5 Preview leads the intelligence index 28.5 to 25.3. Benchmarks, sample bills, speed and a verdict.

Pick it for Mistral Large 4 Preview

  • Lower cost: $1.12 against $1.54 for 1 million input and 200,000 output tokens.

Pick it for Step 5 Preview

  • Higher intelligence index: 28.5 against 25.3.
  • Shorter wait for the first token: 0.91 seconds against 1.9 seconds.
  • Faster output: 151 tokens per second against 130 tokens per second.

By the numbers

MetricMistral Large 4 PreviewStep 5 Preview
Intelligence Index25.328.5
Input price (per 1M tokens)$0.70$1.00
Output price (per 1M tokens)$2.10$2.70
Output speed (tokens/s)130151
Time to first token1.9s0.91s
Release date2026-10-062026-09-18

Mistral Large 4 Preview (Mistral) and Step 5 Preview (StepFun) meet here on their recorded benchmark, price and speed data. The figures come from tracked catalogue data, not from our own testing, so treat them as a starting point for a trial on your tasks. The headline gap is on the intelligence index, where Step 5 Preview scores 28.5 and Mistral Large 4 Preview 25.3.

How the intelligence index compares

The catalogue puts Mistral Large 4 Preview at 25.3 and Step 5 Preview at 28.5 on its intelligence index. A lead of this size is hard to dismiss as noise.

Among 109 active scored models, Mistral Large 4 Preview sits 67th and Step 5 Preview 61st.

Specialist coding and math indexes are missing for both, so treat the overall score as the only capability signal here.

Release timing

The release dates are September 18, 2026 for Step 5 Preview and October 6, 2026 for Mistral Large 4 Preview, 18 days apart. That is close enough to count them as the same generation.

Step 5 Preview is the faster writer

Once a response is under way, Mistral Large 4 Preview produces roughly 130 tokens per second against 151 tokens per second for Step 5 Preview. On long outputs, that is a real advantage for Step 5 Preview. The wait for the first token averages 1.9 seconds on Mistral Large 4 Preview against 0.91 seconds on Step 5 Preview. Mistral Large 4 Preview waits 2.1 times as long before its first token.

For an 800-token response, first token plus generation comes to roughly 8.1 seconds on Mistral Large 4 Preview and 6.2 seconds on Step 5 Preview.

What they cost

Per million tokens, Mistral Large 4 Preview charges $0.70 for input and $2.10 for output, while Step 5 Preview charges $1.00 and $2.70. For output tokens, Step 5 Preview is 1.3 times the price of Mistral Large 4 Preview.

Take 1 million input tokens and 200,000 output tokens: that comes to $1.12 on Mistral Large 4 Preview and $1.54 on Step 5 Preview. For a month of coding-assistant traffic at 30 million input and 3 million output tokens, Mistral Large 4 Preview would come in 28% lower: $27.30 for Mistral Large 4 Preview against $38.10 for Step 5 Preview. Blended at three input tokens per output token, the effective rates are $1.05 for Mistral Large 4 Preview and $1.425 for Step 5 Preview.

Score per dollar

Dividing the blended price by the intelligence index gives a rough cost per point: $0.042 for Mistral Large 4 Preview and $0.050 for Step 5 Preview, per million tokens. Mistral Large 4 Preview delivers each index point for 17% less; whether Step 5 Preview's extra 3.2 points are worth the premium depends on how often your tasks need them.

FAQ

Is Mistral Large 4 Preview better than Step 5 Preview?

Neither wins everywhere: Step 5 Preview scores higher on the intelligence index, 28.5 to 25.3; Mistral Large 4 Preview is 26% cheaper on blended price; and Step 5 Preview streams output 1.2 times as fast.

Which model has lower latency?

Step 5 Preview generates output faster, at 151 tokens per second against 130 tokens per second. Step 5 Preview reaches its first token sooner (0.91 seconds against 1.9 seconds).

Who makes Mistral Large 4 Preview and Step 5 Preview?

Mistral Large 4 Preview is made by Mistral and Step 5 Preview by StepFun. Mistral Large 4 Preview was released on October 6, 2026 and Step 5 Preview on September 18, 2026.

When were Mistral Large 4 Preview and Step 5 Preview released?

Mistral Large 4 Preview is newer. It was released on October 6, 2026, 18 days after Step 5 Preview (September 18, 2026).

Is Step 5 Preview cheaper than Mistral Large 4 Preview?

Mistral Large 4 Preview is cheaper on a blended basis. Mistral Large 4 Preview lists $0.70 per million input tokens and $2.10 per million output tokens; Step 5 Preview lists $1.00 and $2.70. For 1 million input and 200,000 output tokens that is $1.12 against $1.54.

Final whistle

The verdict: Step 5 Preview has more going for it here, but Mistral Large 4 Preview is not without an argument. Pick Step 5 Preview if you need the higher overall score (28.5 against 25.3) and users are waiting on the first word. Pick Mistral Large 4 Preview if cost matters ($2.10 against $2.70 per million output tokens).

Where the numbers come from

The scores, speed and prices in this comparison come from independent third-party model evaluations that Franklin AI syncs each day, and the article was written from a snapshot of them. For the latest results, open the Mistral Large 4 Preview and Step 5 Preview pages. The AI Benchmark methodology explains how that data is collected and shown.

Comments (0)

No comments yet

Be the first to share your thoughts!