Pricing Update April 2026 — Aligning with Real Upstream Costs
1vy synced its pricing for 13 models to match current upstream supplier rates (Anthropic, OpenAI, Google, DeepSeek, xAI). Most models got cheaper. Two got more expensive. Full transparency below.
Why this update
When 1vy first launched, prices for several flagship models were set as placeholders — the migration that seeded them carried a comment saying "verify before launch." During an audit on April 18, 2026, we cross-checked every catalog entry against each provider's official pricing page and corrected the gaps. This post covers exactly what changed.
1vy's pricing policy has not changed: flat 15% markup over the upstream supplier list price. What changed is that some entries were not at 15% — some were higher (overcharging), some lower (a loss). All have been brought into the policy.
TL;DR
| Provider | Models affected | Direction | Magnitude |
|---|---|---|---|
| Anthropic | claude-opus-4-6 | ⬇️ cheaper | -67% |
| Anthropic | claude-haiku-4-5 | ⬆️ more expensive | +25% |
| gemini-2.5-flash | ⬆️ more expensive | +700% (output) | |
| OpenAI | gpt-5 | ⬇️ cheaper | -75% (input) |
| OpenAI | gpt-5-mini | mixed | input -38%, output +25% |
| OpenAI | gpt-5-nano | ⬆️ more expensive | +100% (output) |
| DeepSeek | all 4 models | ⬇️ cheaper | -60-80% (output) |
| xAI | grok-4 | ⬇️ cheaper | -60% (output) |
Plus 8 net-new models added: claude-opus-4-7, claude-opus-4-5, claude-opus-4-1, claude-sonnet-4-5, gemini-3-flash, gemini-2.5-flash-lite, kimi-k2, kimi-k2.5.
What got cheaper (and why)
Anthropic claude-opus-4-6
Was priced as if it were the original Opus 4 ($15/$75 supplier). Anthropic actually dropped Opus 4.6 to $5/$25 at launch. Customers were paying 3.45× real cost on this model.
| Old 1vy price | New 1vy price | |
|---|---|---|
| Input | $17.25 / 1M tokens | $5.75 / 1M tokens |
| Output | $86.25 / 1M tokens | $28.75 / 1M tokens |
If you've been using Opus 4.6 heavily, your monthly bill for this model just dropped substantially.
OpenAI GPT-5 family
GPT-5 was seeded with $2.50 supplier input — actual is $1.25.
| Model | Input change | Output change |
|---|---|---|
| gpt-5 | $5.75 → $1.4375 (-75%) | $23.00 → $11.50 (-50%) |
| gpt-5-mini | $0.46 → $0.2875 (-38%) | $1.84 → $2.30 (+25%) |
| gpt-5-nano | unchanged | $0.23 → $0.46 (+100%) |
DeepSeek (V3, V3.2, R1, chat, reasoner)
DeepSeek significantly reduced its API rates over the past several months. 1vy had old rates baked in — sometimes 5-6× current upstream cost.
All four DeepSeek-aliased models are now at: $0.322 input / $0.483 output per 1M tokens.
xAI grok-4
Was priced at $3.45/$17.25. Actual upstream is $2.00/$6.00.
New 1vy: $2.30 input / $6.90 output per 1M tokens.
What got more expensive (and why)
Google gemini-2.5-flash
This one stings. It was seeded as if Gemini 2.5 Flash carried Gemini 1.5 Flash pricing ($0.075/$0.30 supplier). Google actually launched 2.5 Flash at $0.30/$2.50 supplier — ~4× input, ~8× output.
| Old 1vy price | New 1vy price | |
|---|---|---|
| Input | $0.0863 / 1M tokens | $0.345 / 1M tokens |
| Output | $0.345 / 1M tokens | $2.875 / 1M tokens |
We were selling this model below cost. Continuing would have meant losing money on every call. If you use 2.5 Flash heavily and need to manage cost, gemini-2.5-flash-lite at $0.115/$0.46 is an excellent alternative — same family, ~6× cheaper output.
Anthropic claude-haiku-4-5
Same story, smaller magnitude. Seeded at $0.80 supplier (extrapolated from Haiku 3.5); actual is $1.00.
New 1vy: $1.15 input / $5.75 output per 1M tokens (was $0.92/$4.60).
OpenAI gpt-5-nano output
Output was $0.23, should have been $0.46. Doubled to match supplier.
What was added
Eight new models you can now use:
- Anthropic: claude-opus-4-7, claude-opus-4-5, claude-opus-4-1, claude-sonnet-4-5 - Google: gemini-3-flash, gemini-2.5-flash-lite - Moonshot: kimi-k2, kimi-k2.5
All available through your existing 1vy API key — just change the model parameter in your SDK call.
Verification
Every change was cross-checked against:
- Anthropic: https://platform.claude.com/docs/en/docs/about-claude/pricing - Google: https://ai.google.dev/gemini-api/docs/pricing - DeepSeek: https://api-docs.deepseek.com/quick_start/pricing - xAI: https://docs.x.ai/docs/models - OpenAI: helicone.ai/llm-cost (OpenAI's own page returns 403 to programmatic fetches) - Moonshot: https://platform.kimi.com/docs/pricing/{chat-k25,chat-k2}
The full audit is documented in commits c1e6640 through d26ace4 on the public repo.
What's still pending
Three providers couldn't be programmatically verified because their pricing pages are SPA-rendered or login-gated:
- Zhipu (GLM): glm-5, glm-5.1, glm-4.6 are out, but pricing page didn't return data. glm-4-plus left at current rate; new versions not yet added. - Baidu (ERNIE): 4.5 / X1 are out, but doc URL returned 404. ernie-4.0 left at current rate. - Alibaba (Qwen): tiered pricing across token ranges, doesn't fit 1vy's single-rate model cleanly. Current Qwen prices left untouched pending decision on whether to expose tiered rates.
If any of these matter for your workload, please reach out — we'll prioritize verification.
Questions
If you're affected by any of these changes — especially the increases on gemini-2.5-flash or claude-haiku-4-5 — and want guidance on alternatives, contact us at [email protected].
For full pricing across all 51 models, see 1vy.ai/models.