What's the cheapest way to run frontier-class coding in August 2026?

DeepSeek V4 Flash (0731) at $0.14 input / $0.28 output per MTok is the cheapest hosted frontier-adjacent coding API as of August 2026 — a full benchmark suite reportedly cost about three cents. Qwen3.8-Max is a 2.4T open-weight model with stronger claimed benchmarks, but self-hosting it needs heavy GPU memory, so DeepSeek V4 Flash wins on pure cost per task today.

Is Qwen3.8-Max better than DeepSeek V4 Flash for coding?

On Alibaba's claimed benchmarks, yes — Qwen3.8-Max reports SWE-bench Pro 67.7 and Terminal-Bench 2.1 86.6, well above a lightweight flash model. But DeepSeek V4 Flash is ~50x cheaper per token via API and needs no GPUs of your own. Choose Qwen3.8-Max for top capability (once weights ship and are verified); DeepSeek V4 Flash for cheapest bulk coding.

How much does DeepSeek V4 Flash cost vs alternatives?

DeepSeek V4 Flash is $0.14/$0.28 per MTok (cache-hit input $0.0028), with a pending 2x peak-hour policy. That's roughly 99% cheaper than GPT-5.5's $5/$30 and far below Claude Sonnet 5's intro $2/$10. Qwen3.8-Max API pricing was TBD at its August 3, 2026 launch.

Quick Answer

DeepSeek V4 Flash vs Qwen3.8-Max (Aug 2026)

Published: August 5, 2026

The Short Answer

For the cheapest hosted coding in August 2026, DeepSeek V4 Flash ($0.14/$0.28) is unbeatable — a whole benchmark run costs pennies. Qwen3.8-Max is the stronger (claimed) 2.4T open-weight model, but self-hosting it is expensive. Cheap bulk coding → V4 Flash. Top capability + open weights → Qwen3.8-Max.

Side-by-Side

	DeepSeek V4 Flash (0731)	Qwen3.8-Max
Vendor	DeepSeek	Alibaba
Type	Hosted (open weights exist)	Open-weight (dropping ~Aug 10)
API price (in/out per MTok)	$0.14 / $0.28	TBD
Cache-hit input	$0.0028	—
SWE-bench Pro*	— (flash tier)	67.7
Terminal-Bench 2.1*	—	86.6
Launched	Jul 31, 2026	Aug 3, 2026

*Qwen3.8-Max figures are Alibaba-reported, unverified independently as of Aug 5, 2026.

How To Choose

Cheapest cost per task → DeepSeek V4 Flash. At $0.14/$0.28 with near-free cache hits, it’s the price-war winner for high-volume coding, agents, and CI.
Highest claimed capability + open weights → Qwen3.8-Max, once weights ship and independent SWE-bench numbers confirm Alibaba’s claims.
Middle ground → DeepSeek V4 Pro ($0.435/$0.87 off-peak) if you want more headroom than Flash without frontier API prices.

The Trade-Off

Cost vs ceiling: V4 Flash is dirt cheap but a lightweight tier; Qwen3.8-Max aims for frontier agentic performance at a much higher compute cost to self-host.
Hosted vs self-host: V4 Flash is one API key. Qwen3.8-Max at 2.4T MoE needs serious GPU memory — “open weight” ≠ free to run.
Peak pricing: DeepSeek has signaled a 2x peak-hour policy (Beijing windows); factor that into budgets.

Watch Outs

deepseek-chat/-reasoner model names retired July 24, 2026 — use current V4 identifiers.
Qwen3.8-Max weights + license unconfirmed at launch — verify before planning a self-host.

Verdict

Cheapest coding today → DeepSeek V4 Flash
Strongest open-weight coding (pending verification) → Qwen3.8-Max

Sources

The Next Web — DeepSeek V4 Flash cheapest model: thenextweb.com
MarkTechPost — Qwen3.8-Max release: marktechpost.com
DeepSeek — pricing: api-docs.deepseek.com