Yapay Zeka
Fiyat Güncellemeleri

We're shipping the AI Model Leaderboard — 6 ranking categories (Value, Speed, Coding, Reasoning, Creative, Vision), live score bars, and direct profile links. Bookmark it. It updates as models ship.

New Savings Calculator lets you pick any two models, set your request volume, and instantly see monthly/annual savings. The shareable URL means your colleagues can validate the math.

The Stack Optimizer lets you select up to 4 use cases and get the optimal model for each. It also compares using a dedicated per-task stack vs. a single model for everything — so you can see which approach is actually cheaper.

Anthropic releases Claude Opus 4.6 at $5/1M input, $25/1M output. Tops every benchmark we track: coding 100, reasoning 98, creative 96. 200K context window with 50% batch discount.

OpenAI ships GPT-5.2 at $1.75/1M input, $14/1M output. 256K context, speed score 85. Strong across coding and reasoning while remaining cost-accessible for production workloads.

DeepSeek slashes V3 input pricing from $0.27 to $0.07 per million tokens — an 74% reduction. Output holds at $1.10/1M. Now one of the best value models for reasoning tasks at scale.

Google releases Gemini 2.0 Flash with a speed score of 95 — the fastest model we track. At $0.10/$0.40 per million tokens with 1M context, it's the go-to for high-throughput vision and chatbot pipelines.

Google's new thinking variant posts a reasoning score of 96 while maintaining Flash-tier pricing. Best-in-class for structured analysis pipelines that can't afford Claude Opus.

Updated SWE-bench Verified scores confirm Claude 3.5 Sonnet at coding: 92, reasoning: 88. Remains the strongest mid-tier model for coding workloads. Compare with GPT-4o for your stack.

OpenAI cuts GPT-4o mini to $0.15 input / $0.60 output per million tokens. With a speed score of 92 and solid reasoning, it's now the default recommendation for budget-conscious chatbot deployments.

Mistral's Large 2 joins the OPTERA database at $2.00/$6.00 per million tokens. Solid reasoning (85) and extraction scores (80) with 128K context. Strong GDPR compliance story for EU deployments.

Meta releases Llama 4 Maverick with coding score 85 and 1M token context. At $0.20/$0.85 per million tokens, it offers serious value for open-weight advocates who need vision capability.