Home / GPT-6 vs Gemini Cost
Gemini 2.5 Pro undercuts GPT-6 Sol on input ($1.25 vs $2.00) but matches it exactly on output ($10.00). The result: a modest gap that comes entirely from the input side — and vanishes on output-heavy workloads.
✓ Last verified: September 26, 2026 · Manually verifiedOne context-length caveat. Google's rates below are short-context: they apply to prompts of at most 200K tokens. Gemini 2.5 Pro charges more for longer prompts, which our calculator does not model — so treat the Gemini column as a lower bound if your prompts routinely exceed 200K tokens. Also note: the Flex/Batch 0.5× tier multipliers in the widget are OpenAI's published tiers, applied uniformly for comparison math — Google runs its own discount programs with different terms, which we don't model. Full policy: methodology page.
Pick a model, set your workload, and watch all four update side by side. Prices come from the same data layer as every calculator on this site — never hard-coded per page.
Per 1M tokens, USD. GPT-6 rates are short-context Standard; Gemini 2.5 Pro rates are Google's short-context (≤200K prompt) rates.
| Model | Input / 1M | Cached input / 1M | Output / 1M | Source |
|---|---|---|---|---|
| GPT-6 Luna | $0.10 | $0.01 | $0.50 | OpenAI |
| GPT-6 Sol | $2.00 | $0.20 | $10.00 | OpenAI |
| GPT-6 Astra | $10.00 | $1.00 | $50.00 | OpenAI |
| Gemini 2.5 Pro | $1.25 | $0.125 | $10.00 |
Note the shape of Gemini's pricing: input at 0.625× Sol, cached input at 0.625× Sol, output at exactly 1× Sol. Any cost difference between the two is therefore 100% an input-side story.
Concrete workload — 30,000 input + 8,000 output tokens per request, 20,000 requests/month, 40% cached input, Standard tier:
| Per request | Per day | Per month | Per year | |
|---|---|---|---|---|
| GPT-6 Luna | $0.00592 | $3.95 | $118.40 | $1,420.80 |
| GPT-6 Sol | $0.11840 | $78.93 | $2,368.00 | $28,416.00 |
| GPT-6 Astra | $0.59200 | $394.67 | $11,840.00 | $142,080.00 |
| Gemini 2.5 Pro | $0.10400 | $69.33 | $2,080.00 | $24,960.00 |
Gemini 2.5 Pro at $2,080/month undercuts Sol ($2,368) by $288/month — about 12%, or $0.0144 per request. Against the other tiers: Gemini is ~17.6× Luna's $118.40 and 82.4% lower than Astra's $11,840. The $288 gap is sensitive to workload shape: it exists only because this workload carries 30K input tokens per request. Shrink the input side and the gap shrinks with it — at zero input tokens, both models bill exactly the same $10.00/1M output rate and the gap is $0.
Good news for comparability: Google's cache discount ratio is 0.1× — identical to OpenAI's. Cached input costs $0.125/1M on Gemini vs $0.20/1M on Sol, the same 0.625× relationship as fresh input. Caching therefore never flips this ranking either.
One subtlety: on the example workload, caching saves Gemini 11.5% ($2,350 → $2,080/month) versus 15.4% for the GPT-6 models. That's not a worse cache discount — it's the same 0.1× — it's arithmetic: output tokens, which can't be cached, are 76.9% of Gemini's bill here (vs 67.6% for Sol), because Gemini's cheaper input shrinks the input side. The cheaper your input gets, the less caching can move the total.
Output-token share of the example bill: Luna 67.6%, Sol 67.6%, Astra 67.6%, Gemini 2.5 Pro 76.9%. Gemini's output share is the highest of the four — a direct consequence of its cheaper input. Two practical reads: (1) on output-heavy workloads (long generations, coding agents), the Sol-vs-Gemini gap compresses toward zero since both charge $10.00/1M output; (2) on input-heavy workloads (massive RAG contexts, document analysis), the gap widens toward the full 0.625× input ratio. Your token mix, not the headline rates, decides which vendor costs less for your workload.
Math only, no verdicts. This page compares cost arithmetic, not model quality. It doesn't claim GPT-6 or Gemini is "better" — and cross-vendor figures are directional estimates: Google's above-200K-token rates aren't modeled, so long-prompt workloads will cost more than the Gemini column shows.
That is simply what the published rates say: Gemini 2.5 Pro input is $1.25 vs Sol's $2.00 per 1M (0.625×), while both charge exactly $10.00 per 1M output. On our example workload the entire $288/month gap comes from the input side.
Google charges a higher rate for Gemini 2.5 Pro prompts above 200K tokens. Our calculator uses the short-context row (prompt at or below 200K tokens), so treat its estimate as a lower bound if your prompts routinely exceed that threshold.
The discount ratio is identical: both discount cached input to 0.1× the standard rate ($0.125 vs $1.25 for Gemini, $0.20 vs $2.00 for Sol). Absolute dollar savings differ only because the base input prices differ.
On purely output-heavy workloads: both charge $10.00/1M output, so the gap converges toward zero as input tokens approach zero. Conversely, input-heavy workloads with large contexts widen the gap in Gemini's favor.
Both are verified on our methodology page — Flash at $0.30/$0.03/$2.50 and Flash-Lite at $0.10/$0.01/$0.40 per 1M — but are not in this page's widget, which focuses on the comparison of the two vendors' highest-priced tiers.
The widget applies OpenAI's 0.5× Flex/Batch multipliers uniformly for comparison math. Google runs its own batch and discount programs with different terms, which we do not model — compare at the Standard tier for the cleanest cross-vendor read.
Independence disclosure: this is an independent cost-estimation tool. It is not affiliated with, sponsored by, or endorsed by OpenAI, Anthropic, or Google. Prices are a snapshot last verified September 26, 2026 and may be outdated — always confirm on the provider's official pricing page before making decisions.