chore(pricing): Update fireworks-ai pricing

[siddharthsambharia-portkey]

🔄 Pricing Update: fireworks-ai

📊 Summary (complete_diff mode)

Change Type	Count
➕ Models added	17
🔄 Models updated (merged)	4

➕ New Models

• deepseek-v3p1
• deepseek-v3p2
• glm-4p7
• glm-5
• gpt-oss-120b
• gpt-oss-20b
• llama-v3p3-70b-instruct
• minimax-m2p1
• minimax-m2p5
• qwen3-8b
• qwen3-vl-30b-a3b-instruct
• qwen3-vl-30b-a3b-thinking
• flux-1-dev-fp8
• flux-1-schnell-fp8
• flux-kontext-pro
• flux-kontext-max
• qwen3-embedding-8b

🔄 Updated Models

• kimi-k2-instruct-0905
• kimi-k2-thinking
• kimi-k2p5
• mixtral-8x22b-instruct

Model → Pricing Category Mapping

Model ID	Category	Input	Output	Notes
deepseek-v3p1	Named: DeepSeek V3 family	$0.56	$1.68	cache read = $0.28 (50%)
deepseek-v3p2	Named: DeepSeek V3 family	$0.56	$1.68	cache read = $0.28 (50%)
glm-4p7	Named: GLM-4.7	$0.60	$2.20	cache read = $0.30 (50%)
glm-5	Named: GLM-5	$1.00	$3.20	cache read = $0.20 (page override)
gpt-oss-120b	Named: gpt-oss-120b	$0.15	$0.60	cache read = $0.075 (50%)
gpt-oss-20b	Named: gpt-oss-20b	$0.07	$0.30	cache read = $0.035 (50%)
kimi-k2-instruct-0905	Named: Kimi K2	$0.60	$2.50	cache read = $0.30 (50%)
kimi-k2-thinking	Named: Kimi K2	$0.60	$2.50	cache read = $0.30 (50%)
kimi-k2p5	Named: Kimi K2.5	$0.60	$3.00	cache read = $0.10 (page override)
llama-v3p3-70b-instruct	Tier: >16B	$0.90	$0.90	cache read = $0.45 (50%)
minimax-m2p1	Named: MiniMax M2	$0.30	$1.20	cache read = $0.03 (page override)
minimax-m2p5	Named: MiniMax M2	$0.30	$1.20	cache read = $0.03 (page override)
mixtral-8x22b-instruct	Tier: MoE 56.1–176B	$1.20	$1.20	cache read = $0.60 (50%)
qwen3-8b	Tier: 4B–16B	$0.20	$0.20	cache read = $0.10 (50%)
qwen3-vl-30b-a3b-instruct	Named: Qwen3 VL 30B A3B	$0.15	$0.60	cache read = $0.075 (50%)
qwen3-vl-30b-a3b-thinking	Named: Qwen3 VL 30B A3B	$0.15	$0.60	cache read = $0.075 (50%)
flux-1-dev-fp8	Image: per-step	—	—	$0.00025/step
flux-1-schnell-fp8	Image: per-step	—	—	$0.00035/step
flux-kontext-pro	Image: per-image	—	—	$0.04/image
flux-kontext-max	Image: per-image	—	—	$0.08/image
qwen3-embedding-8b	Embedding	$0.008	$0	—
qwen3-reranker-8b	SKIPPED	—	—	Reranker — excluded per skill rules

Pricing Rules Applied

• Cache read: 50% of input for tier-based models; page-specified override for GLM-5 ($0.20), Kimi K2.5 ($0.10), MiniMax M2 ($0.03)
• Batch inference: 50% of serverless input + 50% of serverless output for all text/vision models
• No cache_write_price set (Fireworks charges read-only)
• Pricing source: 3 consistent scrapes of https://fireworks.ai/pricing (2026-03-08), verified against live page fetch (2026-03-30, same page size ~280KB)

---

Generated by Pricing Agent on 2026-03-30