🎁 New Sign up free, 10 calls on us. Up to $1, no card needed.

Gemini 3.1 Flash-Lite vs Gemini 3.7 Flash

vs

Which one, when — curated verdict, not a benchmark table

Both models share a 1048576-token context window, 65536-token max output, and the same text, image, video, and audio input with text output, so the split is price against reasoning: gemini-3.1-flash-lite-preview costs $0.25 per million input and $1.5 output, while gemini-3.7-flash charges 3x that on input ($0.75) and 2.5x on output ($3.75) and adds a reasoning capability flag. Pick gemini-3.1-flash-lite-preview for high-volume multimodal work where cost dominates; pick gemini-3.7-flash for tasks needing reasoning, or for repeated prompts, where its $0.075 cache-read rate applies.

Pricing

Gemini 3.1 Flash-Lite Gemini 3.7 Flash Δ
Input / 1M tokens $0.25 $0.75 0.33×
Output / 1M tokens $1.5 $3.75 0.4×
Cache read / 1M tokens $0.075

Rates from the live catalog at build time; each model page carries the current card.

Where they sit — input price per 1M tokens across all 63 chat models on this billing unit (log scale)

Capabilities

Gemini 3.1 Flash-Lite Gemini 3.7 Flash
Tool use yes yes
Thinking control yes — vendor dial not published yes — vendor dial not published
Structured output yes yes
Prompt caching implicit + explicit implicit + explicit
Cache lifetime not published not published
Minimum cached prefix 4096 tokens 4096 tokens

Specs

Gemini 3.1 Flash-Lite Gemini 3.7 Flash
Input modalities text image audio video text image audio video
Output modalities text text
Released 2026-03-03 2026-08-13
Knowledge cutoff 2025-01
Context window 1M 1M
Max output 66K 66K

Specs are transcribed from each vendor’s documentation; a row a vendor does not publish is left out rather than inferred. Full sources: Gemini 3.1 Flash-Lite · Gemini 3.7 Flash

One prompt, both models — measured through the gateway

PROMPT Explain to a CFO, in exactly three sentences, why our LLM bill doubled after we enabled reasoning. CHECK exactly 3 sentences

Gemini 3.1 Flash-Lite passed · 3 sentences

Enabling reasoning models increases your costs because the system now generates a significantly higher volume of "hidden" chain-of-thought tokens required to process complex logic. Since you are billed for every token produced during this internal deliberation phase, your total output volume has expanded alongside the depth of the analysis. You are essentially paying for the model to "think out loud" before arriving at a final answer, which directly scales your operational expenditure.

out 86 tok latency 1.4 s

Gemini 3.7 Flash passed · 3 sentences

Enabling reasoning causes the model to generate a high volume of invisible "thinking tokens" behind the scenes to deliberate and solve complex logic before delivering a final response. Because our vendor bills strictly on token volume, this additional internal computation effectively doubled our billable output per query. In short, our costs increased because we are now paying for the extra compute capacity required to produce significantly higher-accuracy answers.

out 878 tok (+799 thinking) latency 6.9 s

Instruction following (exactly three sentences — countable), audience shaping (CFO register), and the hidden-thinking billing gap the token meters below expose.

PROMPT This function is supposed to return the longest run of consecutive equal items, but callers report it is off by one on single-element inputs and misses runs that end at the last index. Fix it and explain the bug in one paragraph. CHECK fix passes tests

Gemini 3.1 Flash-Lite passed · 8/8 cases

### Fixed Code ```python def longest_run(xs): if not xs: return 0 best = 1 cur = 1 for i in range(1, len(xs)): if xs[i] == xs[i - 1]: cur += 1 best = max(best, cur) else: cur = 1 return best ``` ### Explanation of the Bug The original function failed because it initialized `best` and `cur` to `0` and effectively counted "transitions" (how many times an element matched its predecessor) rather than the total count of elements in a run. By starting at `0`, a single-element list returned `0` instead of `1`, and because `cur` reset to `0` and only incremented when a match was found, the function effectively counted "extra" matches beyond the first element, failing to account for the initial element of any run. […]

out 271 tok latency 2.0 s

Gemini 3.7 Flash passed · 8/8 cases

Here is the corrected function: ```python def longest_run(xs): if not xs: return 0 best = 1 cur = 1 for i in range(1, len(xs)): if xs[i] == xs[i - 1]: cur += 1 else: cur = 1 best = max(best, cur) return best ``` ### Explanation of the Bug The original function tracked equal adjacent transitions instead of the actual number of elements in a run, causing all non-empty run lengths to be undercounted by one (e.g., returning `0` instead of `1` for single-element lists because the loop never executes, and `1` instead of `2` for pairs). Initializing both `best` and `cur` to `1` (after handling empty inputs) and resetting `cur` to `1` on a mismatch ensures every individual element is counted as a valid run of at least length one.

out 1050 tok (+827 thinking) latency 6.4 s

Whether the fix is actually correct (runnable), explanation density, and token efficiency on a bounded task.

PROMPT Extract a JSON object with fields {company, quarter, revenue_usd_m, growth_pct, guidance} from this text. Use null for anything not stated; add no other fields. CHECK valid JSON, schema exact

Gemini 3.1 Flash-Lite passed · 5/5 fields, guidance null

```json { "company": "Meridian Robotics", "quarter": "September", "revenue_usd_m": 412, "growth_pct": null, "guidance": null } ```

out 56 tok latency 1.3 s

Gemini 3.7 Flash passed · 5/5 fields, guidance "withheld"

```json { "company": "Meridian Robotics", "quarter": "September quarter", "revenue_usd_m": 412, "growth_pct": null, "guidance": "withheld" } ```

out 918 tok (+858 thinking) latency 6.2 s

Schema obedience (no invented fields), hallucination pressure (guidance is explicitly withheld), and structured-output path differences.

PROMPT Write a 120-word product announcement for an API gateway feature that caches prompts across providers. Forbidden words: "seamless", "unlock", "game-changer", "revolutionize", "empower". Exactly one sentence must be a question. CHECK 120 words, 0 banned words

Gemini 3.1 Flash-Lite passed · 125 words, 0 banned, 1 question

We are launching a new caching layer for our API gateway that stores LLM responses across multiple providers. By capturing frequently requested prompts, this feature significantly lowers latency and reduces operational costs for your applications. Instead of querying expensive models for repetitive tasks, your system retrieves identical outputs instantly from our high-speed cache. This integration maintains consistency while supporting diverse provider workflows, ensuring your infrastructure remains both efficient and scalable. Are you ready to optimize your token usage and improve response times for every user? Configuration takes only minutes through our existing dashboard. This addition provides a practical strategy to manage API spend without sacrificing performance or quality. […]

out 140 tok latency 2.3 s

Gemini 3.7 Flash passed · 120 words, 0 banned, 1 question

Why pay twice for identical AI queries simply because you routed them to different model vendors? Today, we introduce Universal Prompt Caching directly within our unified API gateway architecture. This capability stores repeated prompt contexts across OpenAI, Anthropic, and local models, instantly returning stored results to eliminate redundant computation fees. When your application sends an LLM request, the gateway inspects the payload, identifies semantic matches, and returns accurate cached responses in under ten milliseconds. Engineering teams can now slash inference latency by eighty percent while dramatically reducing monthly token expenditures across diverse production deployments. You retain complete privacy control, flexible cache eviction policies, and granular metrics through a single dashboard. Update your routing settings today to accelerate overall system performance.

out 2858 tok (+2718 thinking) latency 14.1 s

Constraint obedience (word budget, banned-word list, the single question), style fingerprint, and length control.

Switch between them with one line

Both ids are in every tab below — the highlighted pair of lines is the only edit. Same endpoint, same key, same request shape.

from openai import OpenAI

client = OpenAI(
    base_url="https://synthorai.io/v1",
    api_key="sk-syn-...",
)

resp = client.chat.completions.create(
    model="gemini-3.1-flash-lite-preview",
    # model="gemini-3.7-flash",  # uncomment this line, comment the one above
    messages=[{"role": "user", "content": "Summarize this diff"}],
)
print(resp.choices[0].message.content)

Get an API key →

FAQ

Which is cheaper, Gemini 3.1 Flash-Lite or Gemini 3.7 Flash?

Gemini 3.1 Flash-Lite is cheaper on input / 1m tokens ($0.25 vs $0.75, 3.0× apart). Other rows may point the other way — the table above carries the full card, and real cost depends on your mix.

Can I A/B test Gemini 3.1 Flash-Lite against Gemini 3.7 Flash without two integrations?

Yes. Both are served through the same OpenAI-compatible endpoint with one API key — switching is a one-line model-string change, so you can route a fraction of traffic to each and compare bills directly.

Do Gemini 3.1 Flash-Lite and Gemini 3.7 Flash support prompt caching?

Cache-read pricing is listed for only one of the two on our feed; where a rate is missing, the provider does not price cached reads separately.

Related comparisons

From our measured studies