MakeMyDemo / FrameLang v0.1

Videos. Tokens. Costs.

Source ↗Prompts & logs ↗Language & guardrails ↗

Models: 5
Accepted after recorded attempts: 5/5
Duration: 8 sec
FrameLang orientations: 1 / video

Trusted HTML · Geometry checked every frame, contrast sampled · Frozen raw references · Different prompts, not a controlled cost test · Local Mac · Per-video USD includes all recorded study attempts · ≥ known minimum · — unavailable · Render costs excluded

Five models, side by side

DeepSeek V4.1 Flash

MIT · model card ↗

With FrameLang · trusted HTML v0.1

Rendered

Attempts: 4
Video from attempt: 4
Total tokens · all attempts: 84,754
Recorded inference USD · this video: $0.035904666
Geometry frames checked: 192

Per-attempt stats

DeepSeek V4.1 Flash, With FrameLang · trusted HTML v0.1: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	check rejected	29,742	5753	21,349	1,792	$0.016393776
2	check rejected	18,536	5797	3,683	2,560	$0.00667008
3	check rejected	20,265	5914	5,029	2,560	$0.00758778
4	Rendered	16,211	5824	1,282	2,560	$0.00525303

Without FrameLang · original reference

Rendered

Attempts: 1
Video from attempt: 1
Total tokens · all attempts: 5,047
Recorded inference USD · this video: $0.001536444

Per-attempt stats

DeepSeek V4.1 Flash, Without FrameLang · original reference: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	Rendered	5,047	1101	1,300	2,048	$0.001536444

GLM 5.2

MIT · model card ↗

With FrameLang · trusted HTML v0.1

Rendered

Attempts: 2
Video from attempt: 2
Total tokens · all attempts: ≥28,645
Recorded inference USD · this video: ≥$0.118283
Geometry frames checked: 192

Per-attempt stats

GLM 5.2, With FrameLang · trusted HTML v0.1: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	Harness deadline	—	—	—	—	—
2	Rendered	28,645	3067	22,993	0	$0.118283

Without FrameLang · original reference

Human repair · model failed

Attempts: 3
Video from attempt: —
Total tokens · all attempts: ≥19,387
Recorded inference USD · this video: ≥$0.05538554

Per-attempt stats

GLM 5.2, Without FrameLang · original reference: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	No response	6,461	11	0	2,549	$0.00617254
2	Check rejected	12,926	1490	8,886	10	$0.049213
3	Harness deadline	—	—	—	—	—

GLM 5.3

GLM-5.3 custom · model card ↗

With FrameLang · trusted HTML v0.1

Rendered · prior source replay

Attempts: 1
Video from attempt: 1
Total tokens · all attempts: 9,550
Recorded inference USD · this video: $0.03440132
Geometry frames checked: 192

Per-attempt stats

GLM 5.3, With FrameLang · trusted HTML v0.1: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	Prior source rejected → unchanged replay rendered	9,550	2097	4,918	12	$0.03440132

Without FrameLang · original reference

Human repair · model failed

Attempts: 3
Video from attempt: —
Total tokens · all attempts: ≥21,158
Recorded inference USD · this video: ≥$0.0854338

Per-attempt stats

GLM 5.3, Without FrameLang · original reference: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	Harness deadline	—	—	—	—	—
2	Check rejected	21,158	1585	17,023	10	$0.0854338
3	Harness deadline	—	—	—	—	—

MiMo V2.6 Flash

MIT · model card ↗

With FrameLang · trusted HTML v0.1

Rendered

Attempts: 2
Video from attempt: 2
Total tokens · all attempts: ≥21,776
Recorded inference USD · this video: ≥$0.00573188
Geometry frames checked: 192

Per-attempt stats

MiMo V2.6 Flash, With FrameLang · trusted HTML v0.1: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	Harness deadline	—	—	—	—	—
2	Rendered	21,776	5503	13,663	0	$0.00573188

Without FrameLang · original reference

Rendered

Attempts: 1
Video from attempt: 1
Total tokens · all attempts: 5,784
Recorded inference USD · this video: $0.000904988

Per-attempt stats

MiMo V2.6 Flash, Without FrameLang · original reference: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	Rendered	5,784	3071	118	2,560	$0.000904988

Tencent Hy4 Preview

Apache-2.0 · model card ↗

With FrameLang · trusted HTML v0.1

Rendered

Attempts: 5
Video from attempt: 5
Total tokens · all attempts: 65,456
Recorded inference USD · this video: $0.093474621
Geometry frames checked: 192

Per-attempt stats

Tencent Hy4 Preview, With FrameLang · trusted HTML v0.1: recorded model attempts
#	Result	Tokens	Output	Cache read	USD
1	check rejected	8,212	5631	0	$0.016235685
2	check rejected	14,314	5642	2,560	$0.01931557
3	check rejected	14,561	5638	2,560	$0.0195149
4	check rejected	14,111	5641	2,560	$0.019144601
5	Rendered	14,258	5639	2,560	$0.019263865

Without FrameLang · original reference

Rendered

Attempts: 1
Video from attempt: 1
Total tokens · all attempts: 6,549
Recorded inference USD · this video: $0.010392149

Per-attempt stats

Tencent Hy4 Preview, Without FrameLang · original reference: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	Rendered	6,549	2493	1,468	2,112	$0.010392149

Previous typed FrameLang study

DeepSeek V4.1 Flash

Typed FrameLang · creative v0.1

Checked · every frame

Attempts: 8
Video from attempt: 8
Total tokens · all attempts: 77,744
Recorded inference USD · this video: $0.027879654
Geometry frames checked: 192

Portrait ↗

Per-attempt stats

DeepSeek V4.1 Flash, Typed FrameLang · creative v0.1: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	lint rejected	7,049	997	2,672	1,792	$0.002444976
2	lint rejected	8,132	971	3,622	3,328	$0.002797434
3	Rendered	18,851	921	12,283	3,328	$0.008280234
4	lint rejected	12,370	1575	7,838	1,792	$0.005827926
5	lint rejected	6,845	1582	421	2,944	$0.001495332
6	lint rejected	6,795	1706	71	1,792	$0.001555476
7	check rejected	10,972	1706	4,223	2,944	$0.003881082
8	Rendered · final replay passed	6,730	1657	318	2,048	$0.001597194

Original raw reference

Rendered

Attempts: 1
Video from attempt: 1
Total tokens · all attempts: 5,047
Recorded inference USD · this video: $0.001536444

Per-attempt stats

DeepSeek V4.1 Flash, Original raw reference: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	Rendered	5,047	1101	1,300	2,048	$0.001536444

GLM 5.2

Typed FrameLang · creative v0.1

Checked · every frame

Attempts: 2
Video from attempt: 2
Total tokens · all attempts: 58,910
Recorded inference USD · this video: $0.23754592
Geometry frames checked: 192

Portrait ↗

Per-attempt stats

GLM 5.2, Typed FrameLang · creative v0.1: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	lint rejected	26,354	1723	21,422	12	$0.10631692
2	Rendered · final replay passed	32,556	1570	26,984	10	$0.131229

Original raw reference

Human repair · model failed

Attempts: 3
Video from attempt: —
Total tokens · all attempts: ≥19,387
Recorded inference USD · this video: ≥$0.05538554

Per-attempt stats

GLM 5.2, Original raw reference: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	No response	6,461	11	0	2,549	$0.00617254
2	Check rejected	12,926	1490	8,886	10	$0.049213
3	Harness deadline	—	—	—	—	—

GLM 5.3

Typed FrameLang · creative v0.1

Checked · every frame

Attempts: 7
Video from attempt: 7
Total tokens · all attempts: ≥56,206
Recorded inference USD · this video: ≥$0.17516168
Geometry frames checked: 192

Portrait ↗

Per-attempt stats

GLM 5.3, Typed FrameLang · creative v0.1: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	lint rejected	28,350	872	24,269	12	$0.11509932
2	Harness deadline	—	—	—	—	—
3	Local OpenCode launch failed	—	—	—	—	—
4	lint rejected	6,574	1450	28	8	$0.01362848
5	Rendered	7,103	1471	293	8	$0.01522708
6	check rejected	6,956	1749	188	8	$0.01554028
7	check rejected · final replay passed	7,223	1743	113	12	$0.01566652

Original raw reference

Human repair · model failed

Attempts: 3
Video from attempt: —
Total tokens · all attempts: ≥21,158
Recorded inference USD · this video: ≥$0.0854338

Per-attempt stats

GLM 5.3, Original raw reference: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	Harness deadline	—	—	—	—	—
2	Check rejected	21,158	1585	17,023	10	$0.0854338
3	Harness deadline	—	—	—	—	—

MiMo V2.6 Flash

Typed FrameLang · creative v0.1

Checked · every frame

Attempts: 4
Video from attempt: 4
Total tokens · all attempts: 73,288
Recorded inference USD · this video: $0.016495696
Geometry frames checked: 192

Portrait ↗

Per-attempt stats

MiMo V2.6 Flash, Typed FrameLang · creative v0.1: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	lint rejected	15,704	1581	10,839	0	$0.00393736
2	lint rejected	26,441	2430	20,499	0	$0.0069118
3	check rejected	18,486	4341	5,274	3,264	$0.0034863192
4	check rejected · final replay passed	12,657	4341	251	1,856	$0.0021602168

Original raw reference

Rendered

Attempts: 1
Video from attempt: 1
Total tokens · all attempts: 5,784
Recorded inference USD · this video: $0.000904988

Per-attempt stats

MiMo V2.6 Flash, Original raw reference: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	Rendered	5,784	3071	118	2,560	$0.000904988

Tencent Hy4 Preview

Typed FrameLang · creative v0.1

Checked · every frame

Attempts: 4
Video from attempt: 4
Total tokens · all attempts: 34,861
Recorded inference USD · this video: $0.052583526
Geometry frames checked: 192

Portrait ↗

Per-attempt stats

Tencent Hy4 Preview, Typed FrameLang · creative v0.1: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	Rendered	16,324	1411	11,621	0	$0.03533856
2	lint rejected	5,913	888	0	0	$0.006411738
3	lint rejected	6,251	888	0	0	$0.00669363
4	Rendered · final replay passed	6,373	876	0	3,328	$0.004139598

Original raw reference

Rendered

Attempts: 1
Video from attempt: 1
Total tokens · all attempts: 6,549
Recorded inference USD · this video: $0.010392149

Per-attempt stats

Tencent Hy4 Preview, Original raw reference: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	Rendered	6,549	2493	1,468	2,112	$0.010392149

Original benchmark · constrained FrameLang prompt

DeepSeek V4.1 Flash

Original FrameLang

Rendered

Attempts: 1
Video from attempt: 1
Total tokens · all attempts: 3,885
Recorded inference USD · this video: $0.000764094

Portrait ↗

Per-attempt stats

DeepSeek V4.1 Flash, Original FrameLang: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	Rendered	3,885	577	495	2,048	$0.000764094

Original raw

Rendered

Attempts: 1
Video from attempt: 1
Total tokens · all attempts: 5,047
Recorded inference USD · this video: $0.001536444

Per-attempt stats

DeepSeek V4.1 Flash, Original raw: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	Rendered	5,047	1101	1,300	2,048	$0.001536444

GLM 5.2

Original FrameLang

Rendered

Attempts: 1
Video from attempt: 1
Total tokens · all attempts: 3,400
Recorded inference USD · this video: $0.0069926

Portrait ↗

Per-attempt stats

GLM 5.2, Original FrameLang: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	Rendered	3,400	531	217	10	$0.0069926

Original raw

Human repair · model failed

Attempts: 3
Video from attempt: —
Total tokens · all attempts: ≥19,387
Recorded inference USD · this video: ≥$0.05538554

Per-attempt stats

GLM 5.2, Original raw: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	No response	6,461	11	0	2,549	$0.00617254
2	Check rejected	12,926	1490	8,886	10	$0.049213
3	Harness deadline	—	—	—	—	—

GLM 5.3

Original FrameLang

Rendered

Attempts: 1
Video from attempt: 1
Total tokens · all attempts: 3,291
Recorded inference USD · this video: $0.006513

Portrait ↗

Per-attempt stats

GLM 5.3, Original FrameLang: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	Rendered	3,291	527	112	10	$0.006513

Original raw

Human repair · model failed

Attempts: 3
Video from attempt: —
Total tokens · all attempts: ≥21,158
Recorded inference USD · this video: ≥$0.0854338

Per-attempt stats

GLM 5.3, Original raw: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	Harness deadline	—	—	—	—	—
2	Check rejected	21,158	1585	17,023	10	$0.0854338
3	Harness deadline	—	—	—	—	—

MiMo V2.6 Flash

Original FrameLang

Rendered

Attempts: 1
Video from attempt: 1
Total tokens · all attempts: 3,487
Recorded inference USD · this video: $0.0002298464

Portrait ↗

Per-attempt stats

MiMo V2.6 Flash, Original FrameLang: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	Rendered	3,487	560	229	2,688	$0.0002298464

Original raw

Rendered

Attempts: 1
Video from attempt: 1
Total tokens · all attempts: 5,784
Recorded inference USD · this video: $0.000904988

Per-attempt stats

MiMo V2.6 Flash, Original raw: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	Rendered	5,784	3071	118	2,560	$0.000904988

Tencent Hy4 Preview

Original FrameLang

Rendered

Attempts: 1
Video from attempt: 1
Total tokens · all attempts: 3,752
Recorded inference USD · this video: $0.004874517

Portrait ↗

Per-attempt stats

Tencent Hy4 Preview, Original FrameLang: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	Rendered	3,752	855	192	0	$0.004874517

Original raw

Rendered

Attempts: 1
Video from attempt: 1
Total tokens · all attempts: 6,549
Recorded inference USD · this video: $0.010392149

Per-attempt stats

Tencent Hy4 Preview, Original raw: recorded model attempts
#	Result	Tokens	Output	Reasoning	Cache read	USD
1	Rendered	6,549	2493	1,468	2,112	$0.010392149

Hand-authored language showcase · 11.5 sec